Danny's Weblog
Using Gimp "alpha to logo" filters
The "alpha to logo" filters make it easy to produce complex and impressive results, but various problems kept me from actually using them for a long time. I recently spent a few hours experimenting and now have a usable grasp of how to make them work; I explain the "gotchas" that I found below. I'm running Gimp 2.6.3; if you have installed a later version, I should mention that later versions apparently have some differences from 2.6.3 and earlier versions in the user interface to layers.
1. Almost all of these filters require that the original image be on a layer with transparency. When I think about this, it makes sense: that allows the filter to be sure what is the image and what is empty space. So if you want to experiment with these filters, the first thing you have to do is make a new layer with transparency and create the original image on that layer. Note that the filter will work from only the current layer of the image.
If you forget to create a transparent layer and to make sure it's the current layer, then most of the filters are greyed out in the menu, with no clue as to why.
Just outlining an area with a selection tool does nothing until you cut the selected area to a temporary layer. That is, the alpha to logo filters act on the entire layer.
2. You probably want to wind up with decorated text, but you need to take special care when running these filters on a text layer – see below. For initially experimenting with the filters I suggest making a few squiggles with the paint tool on a 640x400 image, and only start trying out text when you are comfortable with the way the filters work.
A nice thing about experimenting with these filters is that if you do Edit - Undo (ctrl-Z) you immediately revert to the state before running the filter – you don't have to keep clicking undo for every stage during the filter.
3. The "cool metal" filter requires that the image is currently a floating selection, rather than merely on a layer with transparency. It is apparently some sort of bug in the version of that filter included with my version of the Gimp, and the filter with the latest versions of the Gimp has been fixed.
If you are using a version of Gimp with this broken cool metal filter, the way to create such a floating layer is to select the image, copy it to the clipboard, and then paste from the clipboard: immediately after the paste, the pasted image is a floating selection, and is identified as such in the layers dialog. Although it is shown at the top of the layers, when you "anchor" the layer it will merge with the previously current layer, not the top layer.
Note that after the filter has run, *everything else in the file* has been destroyed, and the file has been resized to just barely contain the resulting image.
4. After these filters have run, the results will be on multiple layers. For instance, the desired background created by the filter will be on a separate layer and can easily be deleted to allow the image to be pasted on top of another image. Any other layers of the original file may be covered up by the created background.
5. Handling text: the problem is that when you create a text layer it is normally created as a layer just large enough to contain the text. When you run one of these filters, it leaves a set of layers which are just large enough for the filter output, and the rest of the original image has cropped to the text layer size! I don't know why they work that way. But it certainly means that if you want to add text created via one of these filters to an image, you need to work on a separate file and then paste the results into the main image.
6. Alternatively – and I just realized this! – you can access these filters via File - Create - Logo instead. This will easily create the text as a separate file, as you need, that you can then paste into your main file. The only issue is that you have to rekey the text if you want to try a new filter.
Note that the layer that you paste into has to be large enough for the "logo". You need to select such a layer in the main file before you do the paste. After you do the paste, the "logo" is shown as a floating selection at the top of the layers in the layer dialog, but it will be displayed as it will finally be in the main window, ie behind any images on layers that are above the destination layer.
If the destination layer is too small, then very little or even none of the "logo" will be inserted into the layer.
7. I've included a sample. It was based on a couple of words in the Unicode font, Khmer OS Fasthand. I've listed instructions on how to produce something similar; they assume some familiarity with general Gimp operations.
-1. Create a new image 640x480.
-2. Add a transparent layer called "mytext". I suggest you do this by opening the layers dialog with Window - Dockable dialogs - Layers.
-3. Create text in the desired font at say 100 px.
-4. While still using the text tool, click "path from text".
-5. Open the paths dialog, eg with Window - Dockable dialogs - Paths.
-6. Select the path in the path dialog (it will probably be the only one). Click the square red button at the bottom of the paths dialog: selection from path. This took an amazingly long time to run on my ancient laptop, perhaps 20 seconds. This may be because of something strange with Khmer Unicode.
-7. Go to the layers dialog and make sure the "mytext" layer is selected. Make the original text layer invisible (not mytext).
-8. Click the bucket fill tool and click somewhere in the text selection to fill it with a color. I suggest using black initially.
-9. Click Filters - Alpha to logo - Chrome.
-10. Run the filter at defaults.
-11. This will produce a couple of extra layers: one with a layer mask next to it, and one called "drop shadow". As a result, the text is shown in a nice chrome finish.
-12. To give it a gold look instead of chrome, click in the layers dialog on the text layer created by the filter. Make sure you select the lefthand thumbnail, not the righthand one showing the mask.
-13. Click the bucket fill tool and select color mode, opacity 100%, pattern fill, pine pattern, fill whole selection. This will apply the color of that pattern to the chrome-effect text, leaving it tinted yellowish: adequately gold.
-14. To get the leatherish finish on the background, click the Background layer in the layers dialog, click the bucket tool, select the Leather pattern, Normal mode, 100% opacity, fill whole selection, and click anywhere on the image.
-15. Select Image - Flatten image to get all the layers together. (You may want to save the Gimp source file first.)
-16. Crop the image to size, eg by making a rectangular selection, then clicking Image - Crop to selection.
-17. Save as jpeg.
2009 Nov 27 [ Fri ]Fixing Gimp bug problem in Linux: BABL version too old!
I have Gimp version 2.6.3 installed. I've noticed that the latest versions which can be downloaded have new features that I'm very interested in: in particular, brush rotation and a lot of extra features in the layers dialog.
Unfortunately Ubuntu does not provide updated packages, and I cannot install the latest versions of Ubuntu because the live DVD does not boot on my system. (Apparently that is due to a problem with the latest kernels, and it's possible to boot to the latest Ubuntu using a previous kernel, but I am not eager to see if that causes complex and mysterious problems.)
So I tried installing 2.7, almost the latest version of Gimp, in parallel to my existing Gimp. (I could only find version 2.7.1 via a git server, and I did not want to install git.) The installation process supports this using the "prefix" command. However, the compile process stopped, saying it needed updated versions of intltool, babl and gegl. (It has to go through the whole compile over again each time to tell you.) When I got to the gegl requirement, it again seemed to be only available via git, and when I checked how many dependencies it had I gave up.
So this article is not about how to install gimp from source. If you were hoping to find that, try here: [http://shallowsky.com/linux/gimpbuild.html]
What I *can* tell you about is how to get around an irritating problem resulting from this partial install. Because I had tried to install the new Gimp separately from the old one, the old one was still present. But it refused to run. This was the error message:
BABL version too old! GIMP requires BABL version 0.0.22 or later. Installed BABL version is 0.1.0. Somehow you or your software packager managed to install GIMP with an older BABL version. Please upgrade to BABL version 0.0.22 or later.
This is as stupid as it sounds. Gimp version 2.6.3 has a bug in the routine which compares version numbers. It has been corrected in subsequent versions of Gimp, but that didn't help me: [http://osdir.com/ml/svn-commits-list/2009-05/msg06198.html]
I tried reinstalling BABL 0.0.22 using Synaptic, and then tried reinstalling Gimp 2.6.3, but it made no difference. Something was still telling Gimp that version 0.1.0 was installed.
I could not find a clear description of how programs detect library versions when they load, or how to change it. I did not feel very confident that "make uninstall" would fix it, and this (German) blogger agrees: [http://blog.tetti.de/node/844]
The following exhaustive page was stunningly complex, but gave me an idea: [http://www.linux.org/docs/ldp/howto/Program-Library-HOWTO/shared-libraries.html]
It mentioned that the path to libraries for some reason is set up as a succession of links. When listed /usr/lib I got this:
dannyw@dannyw-muggle:/usr/lib\ 10:20:55 $ ls -l libbabl* -rwxr-xr-x 1 root root 816 2009-11-21 13:52 libbabl-0.0.la lrwxrwxrwx 1 root root 21 2009-11-23 09:31 libbabl-0.0.so → libbabl-0.0.so.0.22.0 lrwxrwxrwx 1 root root 28 2009-11-23 10:19 libbabl-0.0.so.0 → libbabl-0.0.so.0.100.0 -rwxr-xr-x 1 root root 371782 2009-11-21 13:52 libbabl-0.0.so.0.100.0 -rw-r – r – 1 root root 101372 2008-06-17 14:51 libbabl-0.0.so.0.22.0
(Actually, I had previously renamed .100.0 to .100.0.danny and made a backup version of .22.0 as .22.0.danny, but I edited the above log to remove that – it still shows up in the filedates)
dannyw@dannyw-muggle:/usr/lib\ 10:21:06 $ sudo rm libbabl-0.0.so.0
dannyw@dannyw-muggle:/usr/lib\ 10:23:42 $ sudo ln -s libbabl-0.0.so.0{.22.0,}
dannyw@dannyw-muggle:/usr/lib\ 10:23:47 $ ls -l libbabl*
-rwxr-xr-x 1 root root 816 2009-11-21 13:52 libbabl-0.0.la
lrwxrwxrwx 1 root root 21 2009-11-23 09:31 libbabl-0.0.so → libbabl-0.0.so.0.22.0
lrwxrwxrwx 1 root root 21 2009-11-23 10:23 libbabl-0.0.so.0 → libbabl-0.0.so.0.22.0
-rwxr-xr-x 1 root root 371782 2009-11-21 13:52 libbabl-0.0.so.0.100.0
-rw-r- -r- - 1 root root 101372 2008-06-17 14:51 libbabl-0.0.so.0.22.0
dannyw@dannyw-muggle:/usr/lib\ 10:23:52 $
The critical thing is that now both libbabl-0.0.so and libbabl-0.0.so.0 point to the .22.0 version. The description of how the system sets up these symbolic links makes it sound as though it may get recreated on reboot, but so far that has not happened. [20091202: the bug reappeared, triggered apparently not by a reboot but by accepting a software update. I was able to fix it the same way shown above. If it happens frequently I will try to find a more permanent fix.]
I can now run gimp without any complaints about babl. Incidentally, reinstalling Gimp via Synaptic seems to have left all my brushes and filters still accessible – phew.
2009 Aug 24 [ Mon ]Creating a document with multiple page orientations in Open Office
I needed to make a simple booklet. Because I wanted it to be easy to copy, I designed it to use just one side of an A4 sheet, folded into 4 to provide 4 working A6 sides, with the inside 4 wasted.
I planned to create the document as a succession of A4 sheets and reduce to a single sheet later. The design involved two landscape-shaped tables as fill-in forms, each requiring a whole page (to be large enough to easily fill in with a pencil). But the remaining pages didn't need landscape orientation, and I wanted to lay them out as portrait orientation.
I thought I knew how to do this but I actually had to take half an hour to figure this out, so I'm posting this in case it saves anybody else some time.
I thought I could apply a different page format if I first divided the document into sections, but that didn't work. (Maybe that's the trick in MS Office, but not Open Office – I haven't checked.) In Open Office Writer, when I changed section 2 to landscape, it changed section 1 and 3 to landscape.
A webpage suggested it would work if you apply a page style (something MS Word doesn't allow), so I created a portrait and a landscape style, but that didn't work either: again, setting a page style to one section set the same style for the other sections.
Eventually I noticed an option in the Insert break... dialog. At the time you insert a page break, you can select a different page style. So if you have portrait and landscape page styles set up, you can set them at that time.
I don't understand why this does not seem to be doable *after* inserting the page break. I suspect that if you try saving to MS Word the results are disappointing.
Similarly, if OO decides to create a new page in the middle of the document – which was happening to me a lot as I adjusted table sizes – it was very tough to delete the page, even though it was apparently empty. I managed it in the end but I'm still not sure what the trick is.
I used the Java utility, Multivalent - Impose to lay out the four A4 pages on a single sheet, as described in a couple of earlier posts.
Command line used: /usr/lib/jvm/java-6-sun/jre/bin/java -cp /home/dannyw/java/Multivalent20060102.jar tool.pdf.Impose -paper A4 -verbose -margin 5mm -dim 2x2 -layout "1u,4u,2l,3l" /home/dannyw/PDF/form01.pdf
The margins used were adequate with my printer, including the margins that I had allowed on the original A4 pages in Open Office Writer.
Note that the Impose utility *forced* the landscape pages to fit inside the portrait-shaped spaces on the A4 page by rotating them. I had to put in the "-l" (that's an "L" for "Left") spec in the command line for those source pages to make it do what I wanted.
Incidentally, I have realized that when I print PDFs something in the print software chain is causing the page size to be forced to A4. I haven't figured out what it is, especially as that option appears to be turned off. Also, the way it works does not seem to be anything anyone would want, because the result is that the bottom of the page is actually outside the printable area in some cases.
2009 Aug 19 [ Wed ]Irritating problem with Open Office Writer -- workaround
I'm running OO Writer version 2.4.1, but apparently this problem has been around for a long time so it's unlikely to ever get fixed. However, at least I figured out what's going on. I hope people who run into it will find this page and waste less time than I did.
I set up a document with headings and subheadings and wanted to set the headings to automatically show heading numbers (1, 1.1, 1.2, 1.2.1, 2, 2.1 etc). I set to work setting the required heading formats in the styles window. After a few minutes work I was surprised to find that no headings at all showed up in the Navigator.
I couldn't fix it and started the file over. After I cut and pasted all the etxt, including all the headings, into a new blank file, I could see the headings in the Navigator, and I breathed a sigh of relief. I started dinking with heading styles again. After a while I realized that the Navigator was never showing level 3 (that is, headings like 1.2.4). Indeed, if you demoted a level 2, it became a level 4, and if you promoted a level 4, it became a level 2 – bypassing level 3.
After another (long, frustrating) while I noticed that if you went to modify the heading 1 or 2 style, under the Numbering tab the Numbering Style showed *blank*, whereas Heading 3 showed as "Numbering 1". There was no way to set it back to blank.
However, when you click Tools - Outline numbering - Numbering - Level 3 - Paragraph style, the "paragraph style" showed as "(none)", whereas the other levels showed style "Heading 1", "Heading 2" etc.
After I set level 3 to show "Heading 3" in Tools - Outline numbering (and similarly dealt with level 5 which I had also dinked with), the Navigator showed and set the outline levels as expected. And all the heading styles showed *blank* in the Numbering Style dropdown.
This seems to be related to the fact that Open Office has two separate and incompatible systems for handling headings, as discussed here (which the writer of the article calls multi-style versus single-style outlining): [http://www.linuxjournal.com/article/7759]
In other words, when you set the format via one route, it silently zeroes out the other setting. So my advice is, only ever go via Tools - Outline.
However, the writer of the above article does not go into the peculiar behavior I ran into. I cannot imagine why anyone would want this behavior. The fact that the Navigator dialog silently skips a level when you try to promote or demote to a mis-set level suggests that someone in the programming team was aware of the issue and put in a brain-dead workaround.
2009 Jul 11 [ Sat ]How to make booklets using the Java utility Impose from Multivalent
It's much more convenient to carry a printout as a booklet instead of a sheaf of A4 pages, but it's quite clumsy and time-consuming to lay out the pages in the right order and orientation even with a powerful layout program. Btw, this kind of layout process is called "imposition". When you set up a layout so that the output sheet can just be folded to create a set of pages in the right order, the set of pages is called a "signature".
I did something similar a few years ago when I created my "Thaitones" booklet, doing the layout in Corel Draw: [http://www.panix.com/~dannyw/pdfs/tonechart.pdf]
It was very fiddly to do the layout, and a lot of work would need to be re-done to create a similar booklet. I had it in mind to find a better way which would make it easier to experiment.
If you are running Windows, you can buy the Clickbook utility. I bought a copy myself more than ten years ago, although I no longer have it. It doesn't seem to have added many features since then, but I remember it being quite convenient and stable, so I'd still recommend it: [http://www.clickbook.com/]
Clickbook works as a virtual printer. You lay out your document in a word processor for your normal paper size, print from your word processor to the quickbook driver, and then the driver gives you options for reducing and rotating pages to produce various types of booklet and so on. It's very nice, holding your hand with instructions for turning the stack of paper and so forth.
If you're running Linux, or Clickbook doesn't have the features you need, then you can download the PDF utility package Multivalent.jar from Multivalent: [http://multivalent.sourceforge.net/download.html]
You should probably check out the info on their website in addition to this article. I am mainly explaining just the Impose tool here, but there are many additional features in the Multivalent package. I hope my extra explanations save somebody some effort.
There are some limitations; for instance, for some reason links don't work. This is probably not a problem for a file intended for print output.
I'm running Ubuntu. Usually I like to install software via the Synaptic package manager. In this case, however, no Ubuntu/Debian package was available. But it turns out the install is not very difficult, although there were several gotchas which I needed to figure out.
The .jar extension on that file means it's a Java file compiled to bytecode. You can just put the jar file anywhere, but then you need to run the java processor on it in the shell, as in this example:
java -classpath ../Downloads/Multivalent20060102.jar tool.pdf.Impose -version
Unfortunately, although the "-version" option to the Impose program seemed to work OK, almost everything else seemed to fail with this error:
java.io.IOException: incorrect header check @ 0
It turns out that Multivalent does not seem to like the Gnu version of Java which I was running:
java -version java version "1.5.0" gij (GNU libgcj) version 4.2.4 (Ubuntu 4.2.4-1ubuntu3)
You would think this would be OK, as Multivalent specifies 1.4, but they also specify Sun as the provider, so I downloaded that:
sudo apt-get install sun-java6-jre
and then everything worked much better. Not only did the utilities work, but also they check the value of the environment variable CLASSPATH; the Multivalent website tells you that you can set that value in the shell so that you don't have to keep setting it as a parameter in the command line. That indeed works with Sun Java, but I could not make it work with Gnu Java.
Still, I hadn't uninstalled Gnu Java, so I needed to explicitly call Sun Java on the command line. When you install it, nothing tells you *where* it was installed, but I eventually found it. You then call it explicitly, like this:
/usr/lib/jvm/java-6-sun/jre/bin/java -version
I would have thought you would also need to tell Sun Java where its libraries were installed, but apparently not, or else it magically works with Gnu libraries.
You can then do something useful like this:
/usr/lib/jvm/java-6-sun/jre/bin/java -cp /home/dannyw/java/Multivalent20060102.jar tool.pdf.Impose /home/dannyw/PDF/danny02.pdf
This takes the input file danny02.pdf and produces an output file in the same directory as the input file, called danny02-up.pdf. The Multivalent site says the output file from running Impose is called danny02-im.pdf, but I never saw that happen. Without any parameters, Impose defaults to a 4-up layout.
I was afraid that a Java utility would be dog-slow, but it was actually gratifyingly fast. Also, the output file was only slightly larger than the input file.
I had a lot of difficulties optimizing the layout for a 16-page booklet on a single A4 sheet. I found the info on the Multivalent website to be a little unclear and misleading. Surprisingly, it was straightforward to make a dummy folded sheet to get the orientations of each page, and set them as required using the "-layout" parameter.
The following works well for me:
/usr/lib/jvm/java-6-sun/jre/bin/java -cp /home/dannyw/java/Multivalent20060102.jar tool.pdf.Impose -paper 210x284mm -verbose -margin 5mm -dim 2x4 -layout "16r,9l,1r,8l,4r,5l,13r,12l,10r,15l,7r,2l,6r,3l,11r,14l" /home/dannyw/PDF/long-test01.pdf
A few notes:
-1. In my original word-processor file, I set *zero* page margins on all sides.
-2. In this example I included the classpath spec, but you should probably set the CLASSPATH variable in your environment as suggested on the Multivalent website.
-3. The syntax for the "-paper" parameter took me a lot of experimentation to figure out. Note that using a "-paper-size" parameter does not work, and that there is no space allowed between "210x284" and "mm".
-4. The Impose utility does not seem to have enough options to absolutely maximize the use of the paper, given that many printers may be like mine in having different unprintable margin sizes on each edge, not to mention varying repeatability in the accuracy with which it locates an image on the page. My own printer seems to have quite good repeatability from side to side, but the main problem was that the unprintable margin was 13 mm on the bottom edge, much larger than the other edges. I found the best way to deal with that was to define a page size that was 13 mm short of the usual 297mm A4 height. You will probably want to experiment to get the command right if you use US Letter size paper.
-5. It is naturally necessary to cut the 13 mm spare edge off the bottom of the output page before folding; otherwise the fold lines will be slightly or significantly off.
-6. Given the various constraints, the Impose utility cannot arrange the source pages on the output page without some wasted space. It appears to deal with that with this layout by putting a gap down the middle of the page. When the page is folded, this gap fortunately winds up on the top of every page of the booklet, so it's easy to snip off. If you like, you can deal with this by creating the original document with a page size whose proportions match the proportions available on the output page, instead of using A4 (or letter).
-7. If you fold the output sheet in the wrong order, you will find that the resulting stack of four-page sheets is also in the wrong order, but that's easy to fix.
-8. Surprisingly, Impose seems to do the right thing if you have a document of fewer or more pages than the number specified with the "-layout" parameter. On the other hand, a 16-page booklet is already fairly bulky; I don't think it would make a lot of sense to try to bind multiple signatures together. If you are lucky enough to have an A3 or larger printer available, though, it might be quite nifty for producing say an 80-page book in five signatures.
-9. Naturally, it's worth using some sort of cutting machine instead of just scissors to provide a nice straight edge when you need to cut the folded sides.
-10. Here's a link to a test file I created. It's set up for my printer, so your printer may wind up cutting off one or more edges of the image; but it should give you an idea of how easy it is to produce a booklet by folding, compared with trying to cut pages individually out of a sheet (as you had to do for my "Thaitones" booklet).
The text in the file is just the current contents of Wikipedia for the Khmer Language page, slightly reformatted. I didn't check details: I just wanted to provide an example of what could fit in a booklet. The Khmer script size has wound up probably a bit too small for beginners to read, although it might look clearer on your printer than it does on mine.
The source document is actually 15 pages, so one page, which happens to be at the top left of the first output page, is blank.
[http://www.panix.com/~dannyw/pdfs/khmer-test01-up.pdf]
You need to print the first page of this pdf on one side, then the second page on the other side. Remember to cut off the bottom margin before folding. I can't think of a foolproof description of the sequence to fold it, but try folding the long side in half three times, while keeping page 1 of the booklet on top each time. Finally, cut off the sides at the top and right of page 1.
-11. You will probably need to dink with the parameters for a while to squeeze the maximum booklet size out of your printable page size. I produced a little diagram to give you some idea of how many things you need to deal with:
It's kind of rough, and the size of the areas is somewhat exaggerated for clarity, although the "-margin" parameter actually sets a margin around each side of each source page, which the drawing shows rather badly. Also, the drawing is not intended to show an *ideal* layout.
If I were planning to deal with random source page sizes and printers, I suppose I would write some sort of wrapper which would take all the values, figure out the best compromise and shell out to the Impose utility. If I knew Java, it would be even better to add a full set ofprinter margin commands to the Java code.
The numbers designate source page numbers; there are eight source pages on this side of this sheet, matching my example file, although the dimensions are not intended to match. I believe it is true that in most cases a printer driver will start trying to print an input image at the top left corner, but I think in many cases Adobe Acrobat will try to *center* and rescale an image to fit on the printable page; check for this. There are so many steps involved that I'm not quite sure of what's happening on *my own* machine.
2009 May 04 [ Mon ]
Problems with Ubuntu updates and repository setup
I've been getting a warning icon about system updates on the desktop command bar. "Your information was last updated 17 days ago."
I tried to run update several times, but (after long delays) got error msgs like this:
Failed to fetch [http://gb.archive.ubuntu.com/ubuntu/dists/gutsy-backports/main/binary-i386/Packages.gz] 404 Not Found
I went to System - Admin - Software sources and clicked Choose best server. This (after a minute) was judged to be ftp.ticklers.org (??).
I tried doing an update again, but (after a while) it still missed several files required for gutsy-backports.
I checked Google, and although there was no clear info, it looked like repositories tend to drop support fairly rapidly, even for versions like 8.04 which are supposed to have extended support. When this happens you need to track down a working repository yourself.
I tried adding various repositories, including old-releases.ubuntu.com, but I still got an error message about 404 looking for gb.archive.ubuntu.com – even when I changed back and had "main server" selected in Ubuntu Updates.
I hand-edited (sudo vi) the /etc/apt/sources-list and commented out other servers for gutsy-backports. It started working; I think now the only server for gutsy-backports is:
deb httx://old-releases.ubuntu.com/ubuntu gutsy-backports main universe multiverse restricted
(Change httx to http)
It found 14 updates, many to basic system files like linux-image and linux-libc. I'm not sure if I should have updated menu.lst or not: it looked like it was trying to install linux...24.24. I don't know how clever the updater is and I don't want to lose various setups in menu.lst like the splash screen.
Doing a "check" now results in no error msgs. "Last updated less than one hour ago." "System restart required."
I don't understand why what I did worked. Surely, if the Update Manager can't find a package at one server, it should try another in sources-list by default?
I rebooted. I opened the /boot/grub/menu.lst file and edited the default boot setup to use .24 for linux and initrd. Booting worked OK and dmesg confirmed that .24 was indeed running OK.
I edited menu.lst to add the new system version and then rebooted again. I had forgotten the "uname -a" command, but that confirmed that the new menu selection was working. I copied the new menu.lst to menu.lst.005 so that I can resurrect it if I ever allow an update to rewrite it.
2009 May 02 [ Sat ]Book review -- Notes on Learning Perl (Schwartz, Phoenix, foy) (O'Reilly 5th edn 2008, for Perl 5.10)
I got this book out of the library a few weeks ago and want to retain some notes about it.
I think I first read this book many years ago, in the first edition. I no longer have that copy, but have a vague memory that the text had very whimsical examples which I found irritating. This edition has a non-grating style.
It's sort of embarrassing that so many of these points were actually new to me. Sometimes, like with the uses of the ternary operator, that just means I don't write programs which are as concise as they could be. But with other points, like using paired characters as the delimiters of a regular expression, this can cause real problems. Of course, Perl is notorious for having irregular syntax and side effects. I really like Perl and would rather program in it than anything else, but I have to admit it may be easier if you already speak Klingon.
1. p14 chap 1 Introduction – A simple program – Wonky version numbering – Perl 5.10 requires "use 5.010". Of course, I had run into this before, but I often forget it. I think the book should have sounded at least a little apologetic here. Why can't everybody just use the same format that's needed for a "use" statement? It's particularly bad that some versions like "use 5.8.8" appear to work, lulling you into a false sense of confidence.
2. p36 chap 2 Scalar data – The chomp operator – "except in cases where it changes the meaning to remove them, parentheses are always optional". Am I the only person who thinks this sounds circular?
3. p60 chap 4 Subroutines – Private variables in subroutines – can you put a space after my like "my ($m, $n)" – generally, what do parentheses mean? This is an example of the problems you get with learning Perl because so many language features – like parentheses – are reused to "do what I mean" in different contexts.
4. p87 chap 5 Input and output – Fatal errors with die – "If you don't want the line number and file revealed, make sure the dying words have a newline on the end" – completely wacky and perlish. I didn't know this before, although now I think about it I probably wondered why the die sometimes displayed the extra info and sometimes not.
5. p137 chap 9 Processing text with regular expressions – different delimiters – "But if you use paired characters, which have a left and right variety, you have to use two pairs: one to hold the pattern and one to hold the replacement string." I didn't realize this at all, much less that the two pairs of characters don't have to be the same – eg you could use []{}.
6. p137 chap 9 Processing text with regular expressions – Case shifting – "The \U escape forces what follows to all upper case". I don't remember this (or \L) before 5.010, although I see it works with 5.008008.
7. p 138 chap 9 Processing text with regular expressions – The split operator – it is unexpected to me that split returns empty leading fields but discards empty trailing fields. The book bleats "here's a rule that seems odd at first" – Yup. Actually, it *still* does. I have a feeling this caused big problems for me years ago when I was trying to scan through a data file.
8. p 139 chap 9 Processing text with regular expressions – The split operator – this refers to noncapturing parentheses, which I thought had not yet been introduced but were actually described on p 125 – blush.
9. p140 chap 9 Processing text with regular expressions – m// in list context – there is an expression here which does several things at once and I hope I can remember it when I get a chance to use it:
my($first, $second, $third) = /(\S+) (\S+), (\S+)/;
-1. It declares multiple variables with a single my
-2. It does a regular expression to set multiple variables
-3. It assigns multiple variables to a list
10. p142 chap 9 Processing text with regular expressions – Nongreedy quantifiers – there is an expression:
s#(.*)#$1#g;
which seems to be missing the <BOLD> tags which it needs to work – see the expression below (in the book).
11. p154 chap 10 More control structures – Autoincrement and autodecrement – there is a nifty expression:
$count($_)++ foreach @people;
This grabs each name from a list of people, and increments the hash value for that hash key for each time the loop encounters that name: ie, if "barney" is in the list three times, at the end of the loop $count{"barney"} = 3.
12. p 163 chap 10 More control structures – The ternary operator – there is an example of a nifty way to set a value to correspond to a range using the ternary operator:
my $size =
($width < 10) ? "small" :
($width < 20) ? "medium" :
($width < 50) ? "large" :
"extra-large"; # default
13. p 164 chap 10 More control structures – The value of a short-circuit operator – "the value of a short-circuit operator is the last part evaluated, not just a Boolean value". I did not realize this. Also, I just had to re-read the section to the end of the chapter to figure out if the word forms "or" and "and", not just "&&" and "||", are also called short-circuit operators.
14. p 165 chap 10 More control structures – Logical operators – example of "my" inside a foreach, first described on p 180 with "this is the first time you've seen it".
15. p185 chap 12 File tests – Stacked file test operators – garbled text needs at least a punctuation fix: "This stacked example is the same as the previous example with just a change in syntax, although, it looks like the file tests are reversed, Perl does the file test nearest the file-name first."
16. p188 chap 12 File tests – Bitwise operators – Fig 12-1 is placed at the top of the page instead of within the "Bitwise operators" section where it belongs.
17. p192 Chap 13 – Directory operations – Moving around the directory tree – "Thus, a tilde prefix will not work with chdir." I keep making this mistake.
18. p249 Chap 17 – Some advanced Perl techniques – Trapping errors with eval – In the example:
eval { $barney = $fred / $dino } ;it points out that "eval" is actually an expression not a control structure like "while", so it needs a semicolon at the end. I remember having endless problems with this when I was learning Pascal.
19. p252 chap 17 – Some advanced Perl techniques – Picking items from a list with grep – "While the grep is running, $_ is aliased to one element of the list after another. You've seen this behavior before, in the foreach loop. It's generally a bad idea to modify $_ inside the grep expression because this will damage the original data."
This is an important point and one which deserves to be stressed repeatedly like this. I remember being stunned the first time I noticed it.
20. p256 chap 17 – Some advanced Perl techniques – Slices – there is an example:
my($card_num, $count) = (split /:/)[1, 5]
I just think it's nifty that you can split $_ into an array of scalars, pick out elements 1 and 5, and assign them, and indeed declare with "my", all in a single line.
21. p273 Answers – Chap 7 exercises – "We showed you an 'or' operation (with the vertical bar, |), but we never showed you an 'and' operation. That's because there isn't one in regular expressions." I have often wondered about that but it's nice to see it explicitly.
2009 Mar 23 [ Mon ]Creating world maps
I have been thinking about creating a world map in Khmer for years.
Wikipedia has a world map in editable .svg form: [http://en.wikipedia.org/wiki/File:BlankMap-World6.svg]
If you open it in Evince and try to zoom in, it looks blurry, but actually it's quite accurate, at least down to the level of England or Cambodia, when you open it in Inkscape.
This also led me to some discussion of how to handle that file: [http://commons.wikimedia.org/wiki/File_talk:BlankMap-World6.svg/Documentation]
and some software intended to automate some tasks with it: [http://gsl-nagoya-u.net/appendix/software/worldmap/]
In particular, the above software allows you to convert the projection used for the map so that it's the correct format to wrap onto a sphere.
The SVG format allows you to edit text inside the file with a text editor; gedit worked OK, even though the files were over 1 MB. It also makes it easy to define multiple areas as parts of one country (under a two-letter nation code eg "gb") and set their color using css. However, I could not figure out any clever ways to handle text annotations like the names of countries: I would have liked some sort of structure like
<country><name>gb</name><title code="en">Great Britain</title><title code="kh">ប្រទេសអង់គ្លេស</title></country>
However, that doesn't really work because you need to set the size and position of each title separately to get the most readable layout, especially when you have more than one language to get in the way.
The world map itself does not have *any* text visible. Here is an example map with a ton of text: [http://commons.wikimedia.org/wiki/File:Map_France_1477-fr.svg] As an experiment I added a little Unicode Khmer using gedit, which seemed to work OK. I could not see how to set text size inside the .svg however.
Furthermore, it seems one should somehow associate the text labels with the countries. The above sample file seems to have that, but to be honest I don't actually see how it's done. I'm guessing each label was manually grouped with the country (as well as manually located and sized).
Here is a .php page which edits web-accessible .svg files on the fly: [http://toolserver.org/~nikola/svgtranslate.php]
It seems to work with Khmer Unicode, to the extent I tried it.
Making Ktoon work
On the (very good) Steve Hanov blog I saw a grumble/sneer about the ktoon program having a bad user interface:
[http://gandolf.homelinux.org/blog/index.php?id=34]
and decided to try it out. I installed using Synaptic. The first time I ran it, it wanted me to choose an install directory, but the "next" button did not work. Eventually I realized "finish" works.
It still wasn't usable. It looked like it couldn't find some of its parts, although there were no error messages. I tried allowing it to set the default install directory, and I think that worked a little better, but still not right.
I then found this page:
https://bugs.launchpad.net/ubuntu/+source/ktoon/+bug/188529
which suggested running this (assuming the default install locations I guess):
sudo ln -s /usr/lib/ktoon/plugins /usr/share/ktoon/
This worked fine and the app was then fairly impressive. You do need to open a new project with "File - Project" to get a normal editable window, but that's somewhat standard.
More irritating, but still only a holdup for a minute or two, was that you can edit a shape by clicking the button with balloon text "selection" and then selecting "contour".
Slow thumbnail generation in Nautilus
I had had Nautilus set to generate thumbnails, which was sometimes very slow, especially in the PDF directory, where Nautilus would take 20-30 s to generate all the thumbnails.
I vaguely remembered this could be set with
gksudo gconf-editor &
but the Nautilus options in gconf-editor did not seem quite what I wanted.
You can however set it inside Nautilus itself:
Edit - Preferences - Preview - Other previewable files - Show thumbnails: never
The only problem is that this is a global setting, and sometimes it's nice to see thumbnails.
Here is a good general discussion of thumbnail generation: [http://ubuntuforums.org/archive/index.php/t-388168.html]
Among other things it mentioned that gqview has useful built-in thumbnail handling. In particular: Edit - Thumbnail maintenance - Shared thumbnail cache - Clear cache
You can also *force* thumbnail generation on a folder and subfolders.
...I then found a page which explained that gconf-edit *can* be used to set behavior for different filetypes individually, but under a key folder I had not looked into before: desktop - thumbnailers - application@pdf
This key shows the command line used to create thumbnails, and (separately) allows you to enable or disable it. Hmm. I didn't change the setting in gconf; I left thumbnails turned off completely in Nautilus (because I usually have all the info I need in the filename, and the icons are almost always too small to really see anything useful), but I may go turn them back on and just disable pdfs.
...I just added /usr/bin/gconf-editor to the System - Preferences menu. Why in heck isn't it installed by default??
2009 Jan 31 [ Sat ]Good Slashdot thread on scientific programming
[http://tech.slashdot.org/article.pl?sid=09/01/30/1443203]
I've seen some interesting descriptions of the R programmable analysis package, and I keep thinking I should get used to Eclipse instead of doing all my programming in vi – indeed, it was only a few months ago that I bothered to make vi do syntax checking.
The above thread contains some very mixed feelings about Labview. Clearly it's too expensive to just play with. But it's pretty much the standard for hardware interfacing, and one poster really liked the way you could reconfigure it on the fly without crashing the experiment.
If only Labview wasn't bloated and hard to program for analysis.
2009 Jan 23 [ Fri ]Good Slashdot discussion on fixing Windows slowdowns
Sometimes Windows decides to start running much slower for no apparent reason. What do you do? The following link contains a lot of ideas: [http://ask.slashdot.org/article.pl?sid=09/01/22/2126205]
Several of the suggestions included "install Linux!" but I have found Ubuntu to be even more liable to intermittent catastrophic slowdowns than Windows, and even less responsive when it happens – ctrl-alt-del will usually bring up a window with an option of rebooting, but over the last few months Ubuntu 7.1 and 8.04 have frequently refused to respond.
One of the first suggestions mentions this issue:
Very commonly this happens when a hard drive reverts to PIO mode after Windows decides it has seen a few errors from the drive.
This happened to me several years ago and I was just amazed that Windows just quietly makes that decision and doesn't even inform you, much less give you a way around it. (I forget the details, but I think I was using an external DVDRW drive on a USB port, and because it (or the disks) was flaky Windows decided to slow down even hard drives on that port, although now I think of it that doesn't make much sense.)
On the other hand, one of the links in that Slashdot article was to a very detailed blog entry by Mark Russinovich: [http://blogs.technet.com/markrussinovich/archive/2008/09/24/3126858.aspx] about a slowdown that he investigated on his wife's Vista machine, and he describes some interesting features built into Windows (apparently since NT4) which make sure the user can get control even after a process has gone haywire. I don't know of anything similar in Linux.
2008 Nov 20 [ Thu ]How to recover from Khmer-keyboard-layout password prompt
I was fiddling around with some Khmer text today and while I was poring through a paper dictionary the screensaver kicked in. When I tried to enter the password prompt, Ubuntu still retained the Khmer keyboard layout setting, so my password didn't work.
I suppose it's nice that the password prompt helpfully mentioned that the Kh keyboard layout was selected, but overall this behavior seems suboptimal. It would seem desirable to flip the password prompt back to the default keyboard layout, just like at startup: after all, it has to be possible to enter the password at startup, so it can hardly do any harm to switch out of the current keyboard layout.
...Hmm, that's true for a single-user machine but I suppose if you have a real multi-user server you need to grab the default setting *for that user*, so that Sophiap or Somchai or Dmitri can choose a password in their own language.
Anyway, there is a workaround. The password prompt that you get after the screensaver kicks in does have an option for "Switch user". This leads you to a full login screen like at startup, which lets you switch the keyboard layout. You can log in as yourself – it tells you you're already logged in – and then you come back to the original desktop. Phew!
To make things easier in future, you can go to System prefs - Keyboard - Layouts - Layout options - Layout switching and choose a key or combination which would allow you to flip the layout while you're at the screensaver password prompt. Indeed, a key was enabled in the default setting after I installed the layouts. However, I rapidly turned it off again because it was so irritating when I hit that key by mistake. Perhaps you feel different.
2008 Nov 18 [ Tue ]Fix for nautilus-actions problem in Ubuntu 8.04
Nautilus is the file manager provided by default with the Gnome window manager in Ubuntu. It allows you to define an executable program to open a certain filetype with, but the normal install doesn't give you many options.
However, you can install the nautilus-actions package, which is much more flexible. And it installs the action in the r-click menu for that filetype.
The install process puts a link in System - Preferences - Nautilus actions configuration.
It just didn't seem to work when I tried it. The action that I created in the GUI didn't show up in the r-click menu. There is some discussion on the ubuntu site about a bug which can be fixed by copying a directory, but that tip didn't help either.
It turns out that Nautilus only seems to load the nautilus-actions code when it starts up for the first time, and apparently Nautilus leaves some code running even after you close its visible windows, so to force it to completely restart you need to enter "killall nautilus". Then when you open a new Nautilus window, it should load nautilus-actions, and also show in the relevant r-click menus whatever actions you have created.
After you've done killall once you don't need to do it again: when you create a new action it shows up immediately... Correction! The entries show up in the r-click menu, but they don't do anything until you restart nautilus. Weird.
2008 Nov 15 [ Sat ]Interesting HTML/CSS trick for headings
A heading is of course a block-level element; in other words at the end of the block there is an automatic linefeed. However, you can get rid of that with CSS. And it made me think I might be able to create a heading which is both left and right justified.
After some tinkering, I reached this example:
<pre> <h1 style="background-color: #11cccc;" >Uploads <div style="margin-top: -1.2em; text-align: right; ">2008-11-14-19-20-33 </div> </pre>
I can't put it in my blog directly, but here's a little test page (including the later modification below): [http://www.panix.com/~dannyw/blosxom/data/test25567.html]
It shows that the right-hand side of the heading is right justified, as you can see by adjusting the window width. Sadly, I have not figured out a way of preventing the left and right sides from overwriting each other if the window width is too small.
There is a "min-width" property in CSS, but apparently it is not supported in Internet Explorer.
...Hmm, it kinda works in Firefox like this (this is the version in the link):
<h1 style="background-color: #11cccc; " >Uploads <div style="margin-top: -1.2em; text-align: right; min-width: 18em; ">2008-11-14 -19-20-33 </div> </h1>
It's not very polished – in fact, as you can see from the fiddly little constants, it's a bit of a kludge. Still, I think it's interesting. I don't remember seeing layouts like this. ...Oops,
<div> isn't allowed in <h1> elements
I guess you could fake the h1 element with a suitable div and css, although that would not work well on non-css browsers. Oh well.
2008 Nov 13 [ Thu ]New procmail info at Panix, and Sudoku
About fifteen years ago I needed to set up a .procmailrc file, and it was a bit of a struggle; I think in the end I needed to beg someone to help. Procmail is software which allows you to set up a bunch of rules to handle email automatically; in the .procmailrc file you can define which emails are to be handled by what commands. Procmail can (and usually does) shell out to external programs like perl, so actions can be arbitrarily complex. For instance, panix provides a spam detector that works via procmail.
A few years ago I moved to Panix, and then realized I needed to set up the .procmailrc again. I think at the time I didn't have access to a backup of the previous .procmailrc file, and anyway I needed to make it do something different, so it was another bunch of work. I think this time I managed to figure it out by myself, except for something about the spam handler itself.
Today I wanted to add another feature to my .procmailrc. The span of time was long enough since my previous adventure that I had forgotten most of the (whimsical) syntax, but when I did a Google search one of the first things that popped up was the following page: [http://www.panix.com/help/mail.procmail.html] which as you can see is produced by panix itself.
This page is really excellent. As well as explaining procmail, it actually gives the most detail about Panix's email software that I have ever seen. So it was quite straightforward to debug the change (eleven tries later – blush).
It reminds me of something I soon realized about learning Unix commands: you don't really start understanding them until you've read through *all* the man pages at least once. Then, the next time you read *all* through them, you understand a bit more. Repeat.
I realized today that this is analogous to Sudoku. In some sense, if a Sudoku grid is solvable, then you could start by solving the cell in the top-left corner, then solve the next on the right, and so on. However, doing it that way would be so difficult that probably nobody has ever done it that way. Instead, you check each cell in turn until you find one that is solvable by executing *relatively few* rules – ie the rules you have managed to figure out for yourself – and then move on.
What particularly made me think of this is the issue of case sensitivity. Even now, I occasionally forget that Unix/Linux is case sensitive. But I've noticed that quite a lot of Unix info, even that obviously aimed at rank beginners, does not mention it, or similar issues. And I'm not really complaining. There just really is a lot of stuff you need to know before you start, and in some ways you need to know *everything* before you start.
So the most efficient learning process has to be iterative, like solving a Sudoku puzzle. And I'm starting to think that it may be generally true for all learning, or at least all efficient learning.
2008 Oct 28 [ Tue ]AMQP -- a standard for middleware
Microsoft has plans to embrace AMQP, a middleware standard promoted by Red Hat among others. [http://linux.slashdot.org/linux/08/10/27/1948210.shtml]
I have often vaguely thought about software to administer normal office operations: things like "receive A, wait for X, do task 275, send result to K35 if less than 1000 and to K11 if greater".
Reading the above article I became aware that any real business operates with multiple large applications and needs to stitch them together. The problem is that if each app has to communicate with each other app by itself, then if management wants to make a change in a business procedure they have to reprogram every app involved in that procedure.
I now see that designing the business with a middleware layer hugely reduces the complexity of changing a business procedure. The individual apps can be designed to handle and generate relatively standardized messages instead of having to incorporate your specialized business logic, so for instance you could install an updated package straight from the vendor and it would continue to work (at least in theory).
All your company's wacky business logic would be contained in the middleware layer, and you could make major changes like adding a completely new app without needing to rewrite any of the software for the other apps: the middleware would re-route and re-format any messages as required.
I apologize if this is old hat to everybody who knows what "middleware" means. Before this I had not grasped the complexity issue.
2008 Sep 30 [ Tue ]Web programming -- avoiding the damned "resubmit form?" prompt
When the browser reloads a page after you submit a form with a POST request, the browser has to ask if you want to resubmit it.
The following Slashdot post describes a way around it. I haven't tried it yet, but it looks good. Basically, you change your CGI code so that after your CGI runs your POST request, it returns a redirect to a GET. The result is that the browser holds the GET page that was displayed *before* the post, then the GET page that was displayed *after* the post. So it can happily move back and forth between those two pages without needing to do any prompts.
[http://it.slashdot.org/comments.pl?sid=980243&cid=25201701]
2008 Sep 08 [ Mon ]MS Access 2007 -- differences from Access 2003
I have used Access 2003 quite a bit, but I've never used Access 2007, so I thought I should probably grab a a book on it. I saw "Access 2007 for Starters" (in the Missing Manual series from Pogue Press / O'Reilly) in the library, and read through it. Despite many grumbles (below), it's well-written and worth buying. The only real grumble about the book itself is that it is intended as a companion to "Access 2007: The Missing Manual", which contains a lot of extra information on programming etc, but also explanations of GUI procedures which one might have expected to find in the "starter guide" version.
Below is a list of notes which I made in order by page location in that book, which mainly address issues where the GUI is unclear or misleading, plus the differences between Access 2003 and Access 2007. I made the notes to have a fast reminder in case I need to start using Access 2007 quickly. But first I want to address the most important thing I learned from the book, which in my opinion is a real gotcha for anyone learning Access.
I assume that most people reading this are already somewhat familiar with Access. One of the nice features is that you have not only a GUI for setting up relationships between tables, but also a GUI for defining a "lookup" field in a table, so that the user can fill in that field by picking from another table.
What I had never realized was that the latter feature only works if the source and destination tables do not already have a relationship. The book states this on page 188:
Access is a bit quirky when it comes to lookups. It expects you to add the lookup, and *then* the relationship. (In fact, when you set up a lookup that uses a table, Access creates a relationship *automatically*.)
If you *do* have an existing relationship, you need to *delete* it before setting up a lookup. This seems to be true for 2003 as well, and explains why I occasionally had baffling problems setting up a lookup in the past. Is there anyone who can think of a logical reason why this restriction is necessary?
Here's a link to my previous posting about MS Access 2003: [http://www.panix.com/~dannyw/weblog/Computers/Opsystems/Windows/msaccess01.html]
Here is a nice simple webpage which lists new features in Access 2007:
[http://www.wiseowl.co.uk/ms-training-version/access-2007-training.htm]
Note that most Access info is either not free or is from MS, so finding info on bugs or shortcomings is tough. The following link may help.
[http://www.allenbrowne.com/Access2007.html]
One standard technique the book does not mention is splitting up a database into two separate Access files: the main (large) data tables, and the forms, queries, menus and VBA code. This makes it much easier to provide updates, or multiple versions using the same live data. However, I noticed in checking the procedures in the book using Access 2003, that when I set up the main data file to use a lookup from a child table to a parent, when I subsequently accessed the child table from the external database file the field showed as the bare ID numbers, instead of as the lookup. Perhaps the full version of the book goes into this issue.
Notes on "Missing manual" book by page
1. p 006 I particularly wanted to pick up simple starter stuff about the new "ribbon" GUI for Office 20076 apps, so I don't apologize very much for noting this absurdly newbie tip: Access does not show the ribbon until you create a database.
2. p 027 You can download the "Access Developer's Toolkit" from MS to allow you to customize the Getting Started page. I don't remember such a thing for Access 2003.
3. p 033 One of the weaknesses of the book is that it is not well organized for people who are familiar with, or need to interoperate with, earlier versions of Access. For instance, this page describes how you can save from Access 2007 in formats compatible with older versions, but does not detail what new features are incompatible with those older formats.
4. p 042 I'm sure these keystrokes were the same in 2003, but I was happy to find them here: use F2 to enter edit mode in a cell, and ctrl-" (ctrl-double-quotes) to copy to a field from the previous record.
5. p 044 You can go to Access Options to set a database to automatically compact on close. (I am ashamed to say I usually forget to compact Access databases, eg when I was messing around with image files in a previous posting.)
6. p 048 Both Access 2003 and 2007 have multiple, clunky, hard-to-read security warnings when you open a database. This page makes the point that going with what appears to be the safer option in 2007 will screw up action queries, of which more below.
7. p 056 You can create a shortcut from the desktop to a table by dragging the table's icon to a blank space on the desktop. I haven't checked to see if you can do the same thing for a query or whatever.
8. p 067 The attachment data type is new. It basically provides the image data type I was grumbling about in my previous Access posting, but can also be used to embed other filetypes, eg Word files. (But there doesn't seem to be any way to do anything useful with the contents of the file, eg getting the creator name of the Word file.)
9. p 071 You can open a "zoom box" to edit a cell – eg a memo field – with shift-F2.
10. p 079 The author seems to think that "2008-23-2" is the "international year-month-day standard". I imagine he means "2008-02-23". It would have been nice if he had been more explicit than "always works": eg, does it work in expressions like filters?
11. p 084 You can edit hyperlinks (add a caption) either by r-click and Hyperlink - Edit hyperlink, or by the shortcut ctrl-K.
12. p 084 Here's an example of a new feature – attachments – which may or may not work if you choose an older database file format.
13. p 099 The book strongly urges you to use an arbitrary (system-generated) ID field as the primary key, instead of figuring out which real (natural) data fields can be used as the primary key. I have gotten the impression from examples created by Microsoft that this is what MS really intends, although I have not seen them explicitly recommending it. For once, I am with MS on this one. See "surrogate key" on Wikipedia: [http://en.wikipedia.org/wiki/Surrogate_key]
14. p 103 I like the new feature of alternating row formats on datasheets.
15. p 113 "If you edit values in a sorted column, then Access doesn't reapply the sort." This is an example of the sort of really fundamental information which I think should be in an introductory section of the book: for instance, it's not obvious to me that Access *does* re-sort if you close and re-open the table (or indeed what behavior is desirable).
16. p 118 The "filter by selection" feature can filter by *part of* a field simply by selecting just that part. That would not have been obvious to me without the book.
17. p 133 The book wisely adds a note about another unexpected behavior: you can enter a record with a *required field* that's set to *blank* simply by entering a space: Access allows this because it sees "something" in the field, but then happily truncates the field to blank (zero characters) when it stores it! You can set "allow zero length" to "no" to prevent this.
18. p 138 This should have been obvious to me, but was not till I read the book. If you want to prevent duplicates in a field, you can set the "indexed" property of the field in Design View to "yes [no duplicates]".
Similarly, you can create an index with a combination of fields, and then prevent duplicates of that *combination* of fields. You may want to set "Ignore nulls" to "yes".
19. p 147 Here it describes how to add a user-defined mask to the mask list, but on p 144 it says you need to buy the *full* book to get this info.
20. p 151 The following is a valid range expression: "Between 0 And 100", ie both "Between" and "And" are keywords.
21. p 156 If you want to set a validation rule for *multiple* fields, that is considered a property of the table, so you do it by going to the property sheet of the table, not of the component fields. If necessary, you can set up multiple rules by using the And and Or operators using a single validation-rule box.
22. p 162 This is the location of half of the info about lookups that I emphasized above.
23. p 179 When you have created a relationship diagram, Access asks you if you want to save it. It is really asking about saving the *diagram* itself. The *relationships* remain in the database whether you save the diagram or not.
24. p 186 At least in Access 2007, if you are viewing a parent table as a datasheet, you can just click the little "plus" box at the left of a record to view its child records. If more than one table is linked to the record, Access lets you pick the table. I can't remember ever knowing that, but maybe I just figured it out from the GUI when I needed it.
23. p 188 This is the location of the rest of the info above about lookups.
24. p 189 You add a lookup by the following procedure
-1. Delete any existing relationships between the tables
-2. Open the child table
-3. Select the field for which you want to provide a lookup to a parent table
-4. In "data type", choose "lookup wizard".
You can then follow the GUI to complete the setup. However, note that in most cases you need to select two fields: the ID field which actually links the tables, plus a human-readable field so that the user can make a meaningful choice. The default setting at the lookup setup window is to not display the ID field, which is usually what you want. For instance, usually the user wants to pick a company name, not the company ID number.
25. p 192 When you first create a lookup, the lookup is not *enforced*; the user can still enter data not present in the lookup. However, you can enforce the lookup by going to the relationship GUI, r-clicking on the relationship line, and enabling "enforce referential integrity".
26. p 193 Similarly to sorts (above), if you add a record to a parent table while the child table is open, you need to refresh the lookup to use the new record in the child table. In Access 2007, you do this by choosing Home - Records - Refresh all.
27. p 203 It is not obvious how to print out a relationship diagram. In 2007, you can do this by clicking Relationship tools - Design - Tools - Relationship report.
28. p 219 The default formatting for query fields (eg backgrounds, fonts) comes from the datasheet setup for the underlying table.
29. p 237 To get an outer join, you first create a relationship in the usual way (ie producing an inner join), and then convert it by r-clicking the relationship line. (I think I knew that before, but I can't remember actually doing it.)
30. p260 This is where the book explains in detail what happens if you did the wrong thing faced with those security warnings when you opened the database. The result is you can't run any code or action queries, but the only message you see is "the action or event has been blocked by Disabled Mode".
31. p 289 In order to export a report to eg Word, you need to do a Print Preview: the required option is part of that mode.
32. p 296 You can't change the formatting of the report unless you're in *layout* mode. (I don't remember knowing this before, but I may have just figured it out from the GUI.)
33. p 308 When you create a form to edit a parent table that's linked to child tables, Access automatically provides a two-part form which allows you to edit the child tables.
34. p 323 In filter-by-form mode, there is no button on the GUI which applies the form. Instead, you have to r-click on an unused area of the form and choose Apply.
35. p 325 If you want to re-use a form filter, you can do Sort & filter - Advanced - Save as query. You can then load the query back in any form which uses the same fields.
36. p 342-346 To connect to an MS SQL Server or MySQL server, click the External data tab, click the More button in the Import section (on the left), then pick ODBC. Then choose "create linked table".
2008 Aug 31 [ Sun ]Adding a new function to an Excel right-click menu
A few months ago I posted a short routine that adds a (hopefully) useful function to Excel's right-click menu when you have a range of cells selected: [http://www.panix.com/~dannyw/weblog/Computers/Programming/Excel/filldown01.html]
At that time I did not explain the tricks for adding that function to the right-click menu. This posting lists those tricks. They aren't rocket science, and I don't go into every detail, but I think it's a better overview than most. You should read through the whole thing before you start trying stuff out.
This information is for Excel 2003. It is probably also good for Excel 2000, but is probably rather misleading for Word, and basically irrelevant for Access. I haven't tried Excel 2007, but I understand MS redesigned the whole menu structure for Excel 2007, so it probably won't work.
The following links may also be useful:
Adding menu items via VBA: [http://www.fontstuff.com/vba/vbatut07.htm]
Creating an Excel addin: [http://www.fontstuff.com/vba/vbatut03.htm]
If you are not very familiar with VBA, note that when you switch to the VBA window many of the top menu items do not work the same way. In particular, you can only get VBA help while you are in the VBA editor, not when you are in normal mode (and vice versa). Also, you cannot do "save as" (including "save as addin") while you are inside the VBA window.
1. The standard term in Windows for what I'm calling "right-click menus" is actually "context menus". However, neither term is relevant to VBA. VBA considers such menus to be simply a type of "commandbar", just like the main menu. In both Word and Excel you can assign a macro to a *non-context* menu via the GUI, but you can only assign a new function to a *context* menu via the GUI in Word; in Excel you need to use VBA, because the GUI doesn't let you (funnily enough).
The term "shortcut menu" usually means a menu accessed by a keystroke combination like ctrl-G.
2. For the sake of trying all this out, here's a very simple function: it makes the text in the currently selected cell (or cells) bold. Starting from a new empty workbook, open the VBA editor with Alt-F11, select Insert Module (to create Module1) and enter the following code in Module1:
Sub MakeBold() selection.Font.Bold = "True" End Sub
(When you test this later, remember that the right-click menu while you are actually editing a cell is a different menu from the one which comes up while you have the cell selected; the MakeBold routine will only work in the latter case.)
3. Then the basic code for creating a new right-click menu item to call that routine is like this:
Private Sub Workbook_Open()
Set NewControl = Application.CommandBars("Cell").Controls.Add
With NewControl
.Caption = "Make bold"
.OnAction = "Module1.MakeBold"
End With
End Sub
You need to put this in the "ThisWorkbook" module. You can do that by opening the VB editor (if it's not already open) with Alt-F11, double-clicking on "ThisWorkbook" (under "Microsoft Excel Objects" under "VBAProject" in the Project window) to open the "ThisWorkbook (Code)" window, and selecting "Workbook" on the left-hand dropdown menu (which initially said "General").
Then the effect is: when the workbook is (next) opened, the "make bold" function gets added to the context menu for each cell (or range of cells). And that feature will *still be there* after you close the workbook, on any other workbook, even after you close and re-open Excel! (Excel has stored the link from the menu to the workbook with the actual code, and will reopen the workbook as soon as you try to execute the menu item.)
4. Now there are two snags. One is that in some cases the above code can add the menu item *every time* the workbook is opened, accumulating more and more copies... and it can happen that they never get removed. The other snag is that it's a pain having to add utility code like this to each workbook where you need it.
5. There are several ways to get around the first problem.
Whatever you do, you need to do two things: don't add a menu item which already exists; and don't forget to remove the item when you exit. But it turns out that the best way to do those two things depends on what you do about the second problem above.
6. Your decision on how to install code like this (any kind of VBA code, not just context menus) depends on whether you want to distribute the code. The big problem is macro security. Excel by default (at least in Excel 2003) refuses to load macros, giving a scary error message and a misleading prompt about how to enable macros which just creates confusion. Unless your users are very used to receiving workbooks containing macros (which allows you to use the signing mechanism – Google search for selfcert.exe), it seems to me to be impractical to distribute macros inside a regular workbook.
7. Additionally, even if the macros are for your own use, it is very tedious to have to reinstall multiple functions and features in each new workbook, and especially to update an existing workbook with new or improved code. (And you have to re-run the signing mechanism for each change, however small.)
8. For these reasons, I think the only practical way to deploy Excel macros is by using the "Add-in" feature. An add-in is just a regular workbook which you put in a special directory for special handling; under 2003 the location is c:\Documents and Settings\dannyw\Application Data\Microsoft\Addins. This is the default location which Excel automatically switches to when you do "save as" "Microsoft Excel Add-in (.xla)".
Instead of manually copying the file to the addin folder, you can use the Tools - Addins - dialog, which allows you to "browse" for another location. Once you select an addin in another location, it will ask you if you want to copy the addin to the standard addins. Even if you don't, Excel will remember to load the addin from the nonstandard location the next time it starts up.
To be any use, an add-in has to contain VBA code. Also, there's no point including regular worksheet data in an add-in, unless it's used by the VBA code – for example, boilerplate text with copyright information which macros might add to every saved workbook.
Once the user has trustingly installed the addin, the macro code will always be available: for instance, the function names are all global, so workbooks can use the functions without needing to specify a module name. And without causing intrusive prompts about enabling macros. (Which makes me wonder about what happens if you just haplessly click on an .xla file link inside a browser.)
However, the function names for some reason don't appear in the list when you insert a function via the GUI; you have to know the name already.
Because the function names are global, you can rename the addin if desired: for instance, if you are a system administrator, you can check that everyone has now installed the latest version of an addin by checking that the filename is "utils-23.xla", and if anyone still has "utils-19.xla" you can grumble at them.
Also, you only need to update a function in one place, and then the updated function is automatically available in every workbook.
9. MS suggests using the "Personal Macro Workbook" for this kind of purpose. But there's only *one* of those per user. If you have multiple users, one or two of them will be already using theirs, and they won't be happy if you overwrite it. Even if the macros are only ever going to be used by you, you may want to keep different sets of macros in different files.
10. Your users will inevitably be exchanging workbooks all the time. If they include macros, every user will need to trust every other user, and/or spend time checking that any macros are safe. If all the workbooks use macros from a single addin, users need to trust only one source.
Additionally, Excel includes an optional weak encryption feature to prevent users opening an addin to view/edit it. This may be cool if you are *selling* the addin, but of course it prevents the educated user from checking that your macros are safe.
11. Because an addin is loaded once at startup, and the functions it provides are intended to stay in memory until Excel closes, the risk of adding menu items more than once is lessened. However, it is still worth putting in defensive code (especially while you are figuring all this out). Also, I felt I should provide some code in case you do *not* deploy the code as an addin.
12. You can check whether a menu item already exists with code like this:
Function Check_MakeBold()
Check_MakeBold = False
For Each Myitem In Commandbars("Cell").Controls
If (Myitem.Caption = "Make bold")
Check_MakeBold = True
End If
End For
End Function
If you open the Immediate window, you can run the routine like this:
debug.Print(check_makebold())
The equality test is case-sensitive, although the function name is not.
13. If necessary, you can delete the context menu item when a regular workbook closes like so:
Private Sub Workbook_BeforeClose(Cancel as Boolean)
On Error Resume Next
Application.Commandbars("Cell").Controls("Make Bold").Delete
End Sub
By including the "on error" you don't need to check whether the "Make Bold" item exists before attempting to delete it.
The above routine needs to be entered in the "ThisWorkbook" module, using a similar approach as for the code running on workbook load (see above).
Another approach is to reset the entire r-click menu to its default state, ie (search Excel Help for "reset method") all the original items and no new ones:
Application.CommandBars("Cell").Reset
This is good if all of your menu additions are created by a single module, but if you have multiple menu items from different modules you will probably want to handle them individually.
Another approach – the one I hinted at in the previous posting – is to include the "temporary" parameter when you add the menu item:
Set newControl = Application.CommandBars("Cell").Controls.Add(, , , , temporary)
14. Once you have the addin installed and running, how do you debug it? It turns out that even the creator of the addin can't open it directly to edit it while it's running as an addin. But if you load it as a normal workbook instead, the functions aren't globally accessible from any workbook.
There is an interesting workaround that I haven't seen described elsewhere: you simply load the file both as an addin at startup, and then as a regular workbook. It seems like that would cause some sort of namespace conflict, but I haven't run into that. I haven't figured out the details of why.
15. If you try all this stuff out, you will accumulate a bunch of functions and want to implement the install procedure by looping through them. I leave that as an exercise for the reader.
2008 Aug 28 [ Thu ]Some grumbles about Ubuntu 7.1 Gutsy Gibbon
I installed Ubuntu for the first time a few months ago. I have installed many versions of Linux, but for a long time I had been using Windows for most things, and Linux (mostly Damn Small Linux) basically for security in internet cafes.
A Linux distribution naturally consists mainly of applications, most of which are the same for different distributions. In this post I will try to address issues which are mainly Ubuntu-related.
1. The main reason I picked Ubuntu was that I had the impression that the majority of Linux apps were available for Ubuntu as binaries. That's also why I installed 7.1, not 8.04 which had already been issued at that time: the apps were already ported and debugged to 7.1.
2. Indeed, I have generally been pleased with installing apps under Ubuntu. The system is easy and fast, with few bugs. I mostly (lazily) used the Synaptic GUI, but occasionally used apt-get when a webpage specified a command-line.
3. I am not quite so happy with the update process. I personally hate any app – including the update app – if it insists on running at at startup and hogging the machine till it completes. OK, it wasn't tough to go in and change the setup so that it doesn't run until manually triggered, but actually what I would have preferred is an automatic reminder, plus – even better – a setup which ran the update at very low priority.
4. An even *worse* point about various processes like the update process was that they *completely took over the interface*. I was unable to run the GUI system tools to even find out what was happening. I was also unable to swap out of X Windows and open a console – at least after I had previously tried to alt-tab. I think this is quite serious: how can it *possibly* happen? I have always hated the way MS Windows takes many seconds to respond to a ctrl-alt-del, but this was way worse. Perhaps Red Hat etc are no better.
5. Actually, I was surprised that ctrl-alt-del does not bring up an interface which allows you to kill an amok process. I would think the most urgent need for an unmaskable interrupt is exactly that: things like logging out are just shortcuts.
6. The system does have multiple language support, but it is not that easy to find, and since there are multiple approaches what info there is seems a little muddled. I was able to set up multiple foreign-language keyboards, but it turns out the interface only allows a maximum of four, which doesn't seem to be documented. You will need to install something to indicate the current keyboard state separately: search for "Gnome indicator applet".
7. I was quite surprised that the internal hard drive was reported as SCSI. The drive hardware is certainly IDE. I haven't seen any other reports of this.
8. The initial install (and operation with the live CD) showed several strange little video problems. For instance, the command bar was initially above the area that the mouse could get to. After some guesswork it seems to work quite reliably now, but I don't really know what I did to fix things.
9. My Motorola V360 works about the same under Linux as Windows – ie the software is full of bugs.
10. For fun I installed "desktop drapes", a utility for automatically switching your desktop background at startup. This turned out to have serious bugs and I have now completely deleted it, mainly because it needs Mono, which I am ideologically opposed to: I don't want anything running MS. NET code on my machine. The only other thing needing Mono was some sort of photo sorting utility which I also deleted.
Also, it turned out that the Mono process was the *worst* offender as far as taking over the machine was concerned.
11. I have still not found a way to avoid the chocolate-colored background that covers the screen at startup. It clashes with my desktop colors and I cannot imagine anyone actually preferring it. (The "splash" screen is something else, which only appears for a second or two on my machine.)
12. Overall I think I might have preferred the KDE desktop. A lot of features seem more mature in KDE, like support for sshfs URLs in the file browser.
13. I found it surprisingly tough to install new fonts, at least by following the docs.
14. I was surprised that the firewall was not running initially. This should be part of the setup.
15. For simple backups you may want to check out grsync. There does not seem to be any backup utility in the default setup.
16. The location where apps get installed, both in the filesystem and in the screen menu, is inconsistent. Some apps do not get a link in the menu system at alll. Perhaps this info is stored somewhere by Synaptic: it was quite irritating to have to check it manually.
17. Firefox would occasionally cause heavy, uninterruptable disk activity – initially I assumed Firefox was the culprit in all such cases. Actually, increasing the size of the cache to 250 MB seems ot have fixed it.
18. Setting up to use the Medibuntu depository (for non-open-source drivers) was surprisingly clumsy and poorly-documented. Also, it didn't appear to work after I installed it, but now seems to be working by magic.
19. I have not been able to make my wifi card work, although it is returning info about the local wireless networks (four or five are visible from my location). It may be something strange about trying to run with both a wifi and a hardwired connection. Certainly the user interface for doing so is full of peculiar bugs.
20. I looked all over for how to set the default handler for files by extension. It turns out that Ubuntu provides a utility for setting a default multimedia handler etc – System - Preferences - Preferred applications. However, it doesn't go by file extension. You set that in your file browser, ie Nautilus in my (standard) case. I suppose that makes sense, but it took me a long time to think of it.
Btw, the problem I was trying to fix was .swfs. Totem had been set to handle them by default, but it couldn't open any .swf I tried. Gnash, on the other hand, works quite well.
21. I was unable to install VMWare, and found a bug report that said it had something to do with having the machine offline during the install. I would have thought that should be the normal procedure.
22. Generally, I have been disappointed with the default Ubuntu setups with respect to security. For instance, the default Firefox setup provided with 7.1 leaves all functions enabled. On the other hand, the default Adblock Plus setup seems designed to actively mislead.
2008 Aug 10 [ Sun ]CPAN install didn't work too well
Although I was able to make CPAN work to the extent of downloading new modules which Perl would *attempt* to run, there were *thousands* of error messages during the install which I could not evaluate. (I have *more than a megabyte* of log files from the various install steps.)
And indeed, the module I really wanted – DBD::sqlite, as I am too cheap to pay for MySQL at Panix – crashes as soon as I try to do anything more with it than load it.
It is not easy to find information on how to handle such problems. I found the following page: [http://sial.org/howto/perl/life-with-cpan/debug/] which gives the impression the writer has been through the process, but is not very useful to the newcomer (unless you like finding "vendor-supplied packages" and compiling them from source).
Fortunately, an alternative showed up shortly after I posted the previous article: [http://www.panix.com/~dannyw/weblog/Computers/Programming/Perl/cpan01.html] Panix happened to upgrade Apache to PHP5 along with Sqlite (perhaps related to me figuring out a way I could make PHP5 run even though PHP4 was installed in Apache), so I can do a lot of the things I thought I needed DBD::sqlite for. Heigh ho, suck air.
2008 Jun 16 [ Mon ]
Setting up CPAN on panix with only user privileges
One of the reasons the perl programming language is so great is that there is a huge variety of modules avilable that you can use to add functionality. The CPAN site (http://search.cpan.org) has thousands. But installing them manually can be so clumsy and error-prone that I've found it easier to write my own code.
I had been aware for *years* that the best way to install modules was by using the CPAN module itself. But each time I tried to set it up at previous hosts or panix, I failed. (It works much easier when you are root, eg on my own Linux boxes.)
This time I seem to have succeeded. ("Danny, 1995 called and wants its issues back.")
Apparently the CPAN module *itself* was installed at panix, so I didn't have to do the whole "make install" thing on CPAN itself. However, the first time I tried running CPAN with something like
perl -MCPAN -e 'install Archie'
it apparently recognized that I didn't have any configuration defined and started asking me a bunch of questions that I didn't really know how to answer. But I knew that the *defaults* for CPAN would be to install to *system-wide* directories which I didn't have write access to. I tried using the "PREFIX" switch for the directories, but it didn't quite work. It looked like the ftp get was failing to bring in the source file, perhaps because of an error reported about "MD5 security checks". (Later I realized that impression was false – that message always appears.)
Incidentally, it's good to use a module like Archie as a test when you're setting up CPAN, because you don't want to mess up the links to a module you actually *need*. You can check whether a module is already installed (assuming your general perl setup has not been screwed up in the meantime!) like this:
perl -MArchie -e 1
which will silently succeed or produce an error message if Archie is not available.
Another irritating issue was that the CPAN setup script presents you with a ginormous list of possible mirrors, and the way to navigate and select them is not obvious. Apparently the list is fed to you via the "more" utility. Here are the tricks: when you get to the list, just press space to get another screenful, and repeat enough times to get to the end. Then you can enter one of the mirror addresses. I'm not sure how to add more than one – I wouldn't bother at this point.
Having licked my wounds overnight, I tried again this morning.
The following webpage would probably be a good guide for most people: [http://www.dcc.fc.up.pt/~pbrandao/aulas/0203/AR/modules_inst_cpan.html]
However, actually what I did was manually edit the ~/.cpan/CPAN/MyConfig.pm setup file that had been produced yesterday. This is what I ended up with:
> cat .cpan/CPAN/MyConfig.pm
$CPAN::Config = {
'build_cache' => q[10],
'build_dir' => q[/home/1/dannyw/.cpan/build],
'cpan_home' => q[/home/1/dannyw/.cpan],
'ftp' => q[/usr/bin/ftp],
'ftp_proxy' => q[],
'getcwd' => q[cwd],
'gzip' => q[/usr/bin/gzip],
'http_proxy' => q[],
'inactivity_timeout' => q[0],
'index_expire' => q[1],
'inhibit_startup_message' => q[0],
'keep_source_where' => q[/home/1/dannyw/.cpan/sources],
'lynx' => q[/usr/local/bin/lynx],
'make' => q[/usr/bin/make],
'make_arg' => q[],
'make_install_arg' => q[/home/1/dannyw/myperllib],
'makepl_arg' => q[PREFIX=~/perl5lib/ \_
LIB=~/perl5lib/lib INSTALLMAN1DIR=~/perl5lib/man1 \_
INSTALLMAN3DIR=~/perl5lib/man3],
'ncftp' => q[/usr/local/bin/ncftp],
'no_proxy' => q[],
'pager' => q[/usr/bin/less],
'shell' => q[/usr/local/bin/tcsh],
'tar' => q[/usr/bin/tar],
'unzip' => q[/usr/local/bin/unzip],
'urllist' => [q[ftp://mirror.nyi.net/CPAN/], \_
q[ftp://ftp.wayne.edu/cpan/]],
'wait_list' => [q[wait://ls6.informatik.uni-dortmund.de:1404]],
};
1;
__END__
Note the following:
1. Info on the web suggested that the "~" should work in this file, and I did use it for the "make5pl_arg", but in general, I used a fully qualified path for the home directory.
2. Web info suggested creating the .cpan etc folders manually first, but I guess running the initial CPAN setup script had done that for me
3. The install script probably also guessed the correct locations for the executables like 'pager' for me. I did verify every one manually (with eg "whereis less").
4. It wasn't obvious what syntax to use to allow multiple urls in the urllist, but I guessed right the first time.
5. The INSTALLMAN1DIR value looks like it would break something, but apparently didn't. Maybe it's not essential, because I don't know how it could work.
Now I could run "perl -MCPAN -e 'install Archie' and it seemed to be OK. But I had not fixed the path for modules in my shell, so trying "perl -MArchie -e 1" failed.
The web info provides the following examples for the bash shell:
export PERL5LIB=${PERL5LIB}:~/perl5lib
export MANPATH=~/perl5lib
I am running tcsh, so I knew I had to make some changes to that. But this is the output of the "setenv" command that works:
PERL5LIB=~/perl5lib/lib
having used eg "setenv PERL5LIB perl5lib/lib". This is what I needed in the .tcshrc file:
setenv PERL5LIB ~dannyw/perl5lib/lib
Remember that you can use "source .tcshrc" to run the commands in that file again without having to log out and back in – although some commands may have unwanted results, for example if you append a directory to a path repeatedly.
Note that my version has an extra "/lib" on the end. It didn't work without it. I don't know why the web example apparently did work for them without it.
Incidentally, the default perl at panix is indeed perl5:
This is perl, version 5.005_02 built for i386-netbsd
I don't know what to do if you're running perl4 or 6 (is anybody?)
I'm not sure why the web example tried to bring in an existing value for PERL5LIB. There wasn't one when I checked (perhaps too late) and running eg "perl -MCGI -e 1" still works. – Later: according to this website:
[http://www.perl.com/doc/manual/html/pod/perlrun.html]
the PERL5LIB variable is
A colon-separated list of directories in which to look for Perl library files before looking in the standard library and the current directory.
so it seems to be a *user-defined* variable, so it is normally empty and it doesn't smash any existing definition if I just set it directly as above. Maybe the bash shell doesn't complain like tcsh if the variable is empty when you try to append to it. – Confirmed.
The MANPATH thing just doesn't work for me at all. Doing "perldoc Archie" works fine without it. Trying to append to MANPATH causes an error message. And if I set MANPATH without appending to the existing value, it makes "man bash" (or anything else) fail.
Some webpages say you need to do this: setenv MANPATH $HOME/perl5lib:${MANPATH} but that failed too. The really funny thing is that $MANPATH doesn't seem to be set to anything in the set or setenv lists, but setting it to something seems to screw up "man". I guess setting $MANPATH *completely overrides* the path set up in man.conf.
When I checked, the default search path for man set in /etc/man.conf seems to be really long and complicated. I couldn't see an easy way to get the current value for it into the environment variable. However it turns out that there's a "-m" switch to man which *adds* a search path to the existing (invisible) path, so I put the following line in .tcshrc:
alias man "man -m ~/perl5lib"
and it worked fine, ie both "man Archie" and "man bash" worked.
Now I have to actually try it with a module I need! Maybe I'll write that up later. If you have been following along and want to try it yourself, remember that cgi programs don't get your path from your user environment, so they will need something like
push(@INC, "/usr/lib/bliv/");
as I discuss in an old posting:
[http://www.panix.com/~dannyw/weblog/Computers/Programming/Perl/perltip01.html]
Useful subroutine in Excel: "filldown"
This is not the most complex stuff I've done in Excel but it is all my own code and does something which I think is generally useful, so I thought I'd list it.
It's intended to be linked to a right-click; I leave that out as an exercise for the reader... especially as I now realize that I wrote the code for that part incredibly badly, although it does work. Hint: there is a parameter for this: application.CommandBars("Cell").Controls.add which prevents an option being added multiple times. See one of the books (I forget which) in my post: [http://www.panix.com/~dannyw/weblog/Computers/Programming/Excel/books01.html]
It's intended to make it easier to enter data in a typical Excel table where the same values recur multiple times. For instance, if you are entering vehicle expenses in a company garage, you probably need to enter the expenses for each car multiple times with the same info so that you can easily sort the data with a pivot table. With this sub, you can just leave a blank for repeating fields, and when you're ready you can just right-click on the table and select the sub, and a value will fill down into all the empty cells below it for each record and field. In many cases it will halve the entry time and cut out a lot of errors as well (ie ensure that each repeated value is identical).
Sub filldown()
' dW 2007-12-20 Loops through a range
' if a cell is blank, copy from previous row
' (we hope)
Dim cellnum, cellrow, cellcolumn, rowcount, columncount As Long
If TypeName(Selection) <> "Range" Then
Beep
MsgBox "This command requires a cell/range to be selected."
Exit Sub
End If
If Selection.Areas.Count > 1 Then
Beep
MsgBox "This command only works on a single range."
Exit Sub
End If
rowcount = Selection.Rows.Count
' You can't fill down if less than 2 rows
If rowcount < 2 Then
Exit Sub
End If
'
columncount = Selection.Columns.Count
' MsgBox ("Rows = " & rowcount & " Columns= " & columncount)
'
cellnum = 0
For Each cell In Selection
cellnum = cellnum + 1
' The \ operator does integer division
' First row is row 0
cellrow = (cellnum - 1) \ columncount
' Don't do anything on first row
If cellrow > 0 Then
cellcolumn = (cellnum - 1) Mod columncount
' MsgBox ("Cell: " & cellnum & "Row: " & cellrow & " Col: " & cellcolumn)
' I seem to have defined cell locations starting at 0,
' but Excel has cell x,y starting at 1,1.
' Need to handle cells with formulas separately
If cell.Value = "" Then
If Selection.Cells(cellrow, cellcolumn + 1).HasFormula Then
cell.Formula = Selection.Cells(cellrow, cellcolumn + 1).Formula
Else
cell.Value = Selection.Cells(cellrow, cellcolumn + 1).Value
End If
End If
End If
Next cell
End Sub
When you play with this, remember that Excel resets the undo stack every time you run a macro. If someone wants to fix this code so that it implements its own undo feature, that would be cool.
Some basic setup info for PHP
I haven't looked at PHP for years and needed to read through Sams' "Teach yourself PHP in 24 hours" recently.
Here are some basic tips for setting things up:
1. Assuming your Apache server allows you to use .htaccess files, you can set a bunch of useful stuff here, eg:
AddType application/x-httpd-php .php
For instance, if most of your site is real .html files but you want to put .php files in a certain folder, you can put an .htaccess file in that folder that tells Apache to run the php interpreter on .html files in that folder:
AddType application/x-httpd-php .html
(You don't want to do that in your root folder because it would waste time running the php interpreter on all the ordinary html files.)
2. The phpinfo() command inside php of course produces a nicely-formatted html file with all the setup information for php, and you need to look at it. In particular, you want to check which php version you have and what modules are loaded (irritatingly Panix has a rather old version with few modules compiled in).
Also, it's good to check the "config-file-path". This (assuming you have shell access) allows you to find the actual php.ini file to check any items not shown by the phpinfo() command.
3. If you don't control the server, you can still set php setup variables inside .htaccess. You use the "php_flag" directive to set a php variable and the php_value directive to set a php.ini variable to a value, eg:
php_flag short_open_tag On
php_value include_path ".:/home/corrdev"
4. Even if you can't use .htaccess files, as of php 4.0.5 you can set some .ini values within php pre itself, eg:
ini_set( "include_path", ".:/home/corrdev" );
5. Don't forget the main php website, eg:
[http://www.php.net/manual]
6. Here's the entire contents of a file you can save as test.php:
<?php phpinfo(); ?>
(I know putting that here makes me look like not the greatest programmer in the world, but every time I start looking at php again I find myself having to work out the basics of the syntax all over again.)
7. Here's some code which looks simple – ok, it is simple – but I find myself wondering whether it will work every time I start playing with php again (this is basically from the book's Listing 5.14 Returning to HTML Mode within a pre block):
<?php
$display_prices= true;
if ( $display_prices ) {
?>
<p>Interesting stuff to show in this case</p>
<?php
}
?>
I don't know about you, but if I had to figure out that syntax from scratch it would take a while.
2008 Feb 02 [ Sat ]
The benefits of learning Excel VBA
The following link explains in amusing detail just how unexceptional the Societe Generale trader M. Kerviel is in his skills, or indeed modus operandi: [http://www.theregister.co.uk/2008/01/31/kerviel/]
From what little I know about the situation in such companies, the article is not misleading.
2008 Jan 29 [ Tue ]Some books I've used for Excel VBA programming
Recently I've had to use Excel quite a lot. I used to be quite snooty about it (although I recognized that – after SQL Server – it was Microsoft's most respectable product). Now, having gotten into VBA programming, I have to admit you get a rather solid development environment.
Anyhow, I wanted to put in links to a couple of books I can recommend:
1. Excel 2007 VBA Programmming for Dummies - John Walkenbach
I actually read his version for Excel 2000 first. The two books are much the same. I had a feeling that he caught some errors in the new version, or maybe I just understood stuff more the second time through (I was unable to compare them except by memory).
I don't think Walkenbach ever did a version of this book for 2003. I don't think there was much difference between Excel 2000 and 2003, but there are a lot of changes in 2007 for the new menu ("commandbar") structure. And the 2007 book *only* documents the 2007 code - people expecting the book to be backwards-compatible with 2003 are going to be very disappointed.
But in general, I found the book quite ueseful.
Walkenbach recommends this "blog": [http://www.dailydoesofexcel.com/]
He also recommends reading Microsoft's own newsgroups via Google. Search Google Groups for microsoft.public.excel.programming, as well as several others under microsoft.public.excel.
2. The Absolute Beginner's Guide to VBA
The advantage of this book is it explains how to run multiple Office apps using VBA: for instance, to read your Outlook mailbox (or at least the header fields and a smidgen of the text) into Excel.
Also, it does show various useful examples in non-Excel apps, and I may need to take a shot at Word VBA sometime.
Incidentally, I rather wish the titles of these books weren't so embarrassing. Don't bring them to your client site! I've actually done some fairly complex stuff with Excel, and if clients ever see this blog they'll get the wrong idea. Hmm, then again if they see some of my other posts my goose is cooked anyhow!
2007 Jul 08 [ Sun ]
Scanning and database-based file access
I have always wanted to transfer my bulging paper files to disk, and I've bought two scanners, but only got around to actually doing a lot of scanning just recently.
When I use the term "OCR" below I am referring to all the software features needed to produce readable text. It turns out for instance that almost every document you ever want to scan – forms, magazine stories, invoices, legal documents – is formatted in blocks on the page, not in single lines. So simple OCR, even if it can read the multiple fonts involved, jumbles together words from separate blocks. If the OCR system can handle block text at all, it usually deals with it by converting to MS Word format, which is actually not very good at handling text blocks, so you wind up with a document which is very hard to edit without smashing the block structure.
At the time I bought the first scanner (in the USA), my idea was to run OCR on everything so that the material was searchable. The OCR performance however was unusably bad, and I let the idea languish. Furthermore, it takes a surprisingly long time to scan and OCR each sheet, especially if there is any difference at all between sheets. (You think you can scan everything and OCR it later, but it turns out that you need to fiddle with all the scan settings to make the OCR work.) Another thing that bothered me was that it was clear that to scan everything at reasonable quality was going to fill more than one disk, especially if I scanned everything at top quality to give OCR the best chance.
Well, technology progressed. About a year ago I bought a new scanner. It had much better resolution, speed and contrast range, but the OCR was still unusably bad. I let it sit there for a long time, until I really had to cut down on my paper files.
The following is some conclusions after doing quite a lot of scans over a couple of weeks.
1. Even if OCR never worked, I really wish I had done scanning on my files before leaving the USA. I really don't need any of the physical stuff I have stored there and could have happily let it go and saved the storage fee, but I can't bring myself to discard the paper files (and photos). I could do so if I had human- readable scans.
2. Even if OCR doesn't work well enough to let the optical scan be discarded, it may still be quite useful in allowing a scan to be located. However, this means that the text is separate from the scan, which requires some sort of database system. I don't actually know of a system that works quite like that; the products I've seen seem to want to pretend that their OCR is perfect, and make it clumsy to get to the original scan from the text. So I haven't implemented this yet. Presumably though my scans are of sufficient quality to get something useful from OCR once I find a suitable product.
3. OCR of course only works in one language. (I was able to use OCR on a Thai subtitles file from a DVD, but of course it doesn't work well with subscript and superscript vowels and tone marks: you have to train it to recognize *each combination of consonant and vowel and tone mark separately* – which I found actually do- able. But the situation with Cambodian, where you have subscript *consonants* too, would be far worse.) Hmm... it occurs to me there has to be some sort of mechanism in the OCR I was using to handle superscript accent marks, because it is rated to work on French. Hmmm...
4. I was able to compare several scanning software packages, and none of them really impressed me. I wound up going back to the basic scan utility that came with my scanner. While it doesn't include OCR, it has a lot of convenience features relative to the software on my first scanner. For instance, you can set up multiple scan areas and create multiple files from a single process. (I can't say "single pass" because the software actually tells the scanning head to scan each area separately – it doesn't scan the whole page and then output slices. On the upside, it allows you to have different setups for different areas. On the downside, I've never figured out how it decides which area to do first.)
The only big disadvantage of the software I was using was that there was no option to save as gif. It would have fit the image type much better. I wound up using jpg for everything. Surprisingly, I found that selecting color did not make a big increase in the jpg filesize: perhaps 30% larger on typical documents. Don't forget to use the descreening filter whenever there's a halftone area in the original: even if it the scan looks OK, the output filesize will be bloated – perhaps by a factor of two or three! – if the software has to try and store all those little blotches. (Your software may call this function something else – probably "moire filter".)
I've seen some examples of very compressed scans using .tif. I think one of the standard compressions for tif is fax compression, which of course is optimized for black text on white paper. My sw actually will produce .tifs, but it doesn't seem to have an efficient compression algorithm – it always looks worse than jpg for an equivalent filesize. I just checked and a tif using CCITT compression was 4 times larger than my jpg for no apparent quality gain.
5. The database issue is important for even fairly simple cases, even though I can't find references to it on the web. For instance, suppose you go to the doctor and get a receipt. Do you think you should store it under the name of the doctor, under "receipts", or under "medical"? Or under "2007"? What about your *girlfriend's* receipts? What if the receipt covers procedures on you *and* your girlfriend?
Suppose you're scanning bank statements. Suppose the statement is dated 2006-01-03, for the period 2005-11-22 through 2005-12-21. What date should it go under? Suppose you don't receive the statement till 2006-06-17? Suppose the bank is taken over by a bank with a different name, but you continue to get items with either name for months?
6. What I'm getting at is that a tree structure is not satisfactory for arranging the kind of item which is typically scanned. You need a database system which allows a file to be tagged and located by a combination of data fields.
7. I had previously been using a product called ThumbsPlus. It is intended for handling photos, especially where you have a website with thumbnail pages allowing you to choose the full-size pictures to download. Other products can produce thumbnail layouts like this, but TP has a lot of convenience and automation features. Its only real limitation is it doesn't allow you to create a page template in Dreamweaver or whatever and use it for its created pages. – Oops – actually it *does* have that feature. Wow.
A particular strong point is its handling of metadata. As well as allowing internal fields inside the jpg to be set – EXIF etc – and read in various automated ways, it can also hold user-defined fields in a separate database. (There is nothing to stop you using predefined fields for your own purposes, but of course it's confusing, and may prevent some automated features from working.)
The separation between the actual file and where the metadata is stored is always a little worrying. For instance, if you do a folder move in Windows instead of using the built-in TP file management, TP loses track of all the external metadata. (It does keep track of volumes by name however, and I think you could use that so that if a directory tree was on a different drive letter for some reason, TP could still match up the files to the separately stored metadata. But I haven't actually verified that.)
I'm surprised that this functionality isn't a bigger part of similar products. At any rate, if it's provided it's not a major part of the specs on their websites.
Database features are provided using a Microsoft Access database. However, you can also point TP at any ODBC data source, including SQL Server and MySQL. (I haven't actually tried that yet, but I have been noticing a slowdown on moving between folders with a database over 100 MB in size, and so I may get around to trying one of them – probably MySQL as it uses less resources.)
8. I therefore think that I should have stored scans simply by scan date – ie, created a folder with the date, and then numbered them 20070605-001.jpg, 20070605-002.jpg. I would then need to set data fields on every file so that they would be locatable. (What I actually did was to try and put files in folders and subfolders with meaningful names, but as explained above they don't really fit that structure.)
I haven't actually *done* this yet because I was busy with the scanning. However there's no reason you can't do it subsequent to scanning, assuming your software has reasonable facilities for applying data to groups of files. Another reason for holding off is that I keep thinking there's a more powerful product than ThumbsPlus out there somewhere – for instance, one which integrates database features for image files and text files (or files which have readable text inside them – for instance a Corel Draw image file might contain a reference to "passport.jpg").
9. An alternative way of accessing files multiple ways would be to make each file part of multiple galleries. A "gallery" is a TP feature which allows you to form groups of files which can easily be browsed through or copied to another location, without needing multiple copies, and means that if you modify the original file the version in the gallery now displays the modified file.
This is probably a little clumsier than the database approach, and does not allow combining choices. On the other hand, the database search featurein TP does not allow searches which are much more complex, eg combining ands and nots as in "country eq "thailand" and (city ne "pattaya") ". Presumably you could do this sort of thing by opening the database file inside Access, but I have been too lazy to try.
10. I have just realized that the "IPTC" fields *inside* the jpg can be used to avoid the necessity for user fields stored in a separate database using two tricks.
First: there are a *lot* of fields inside the IPTC area. You can just decide to use them for your own purposes, for instance you could use the "Urgency" field to mean how good the shot was on a range of 0 through 9.
Second: the IPTC editor allows you to copy info back and forth between IPTC fields and other database fields. So you can use separate database fields for normal functions, but intermittently copy them into IPTC so that they are safely "backed up" in case anything goes wrong with the database.
11. Fortunately a TP database, at least when it's still in Access, is just a single (large) file. So it can easily be backed up. I haven't encountered a case of TP database file corruption, and I've been too lazy to do a disaster test (doesn't mean you shouldn't) so I haven't tried using an old version of a database file deliberately, but I've certainly (by mistake) made major changes to the actual image directory structure outside of TP, and when I reload TP there are no apparent problems – it just isn't up to date on folders which have been added or removed (or moved).
12. Now a few notes on the physical side of scanning.
13. On my scanner the bed is (as you might expect) a little larger than the actual scannable area. What that means is that if you align the workpiece with the sides of the bed, you lose part of the image. (Quite often marginal scribbled notes are the most important part of the image.) But if you don't locate the workpiece somehow, it's in a different location every time, and probably way off perpendicular too.
As a cheap way round this, I scotch-taped two plastic rulers to a horizontal and vertical edge of the bed, allowing me to rapidly and repeatably locate the workpiece at an exact location. If I had realized how much easier it would make everything, I would have cut two 30-cm rulers so that they fit exactly at a mitred edge (and fit fully inside the bed instead of slopping over at one end). Ideally, they would be cut down in width too, so that you could still use them to locate a full page. It might be better to try creating both axes at once by cutting an existing right-angled thing, like a drafting triangle, and then gluing ruler sections on top of it.
When I think about it I'm a little surprised that you don't get a plastic gizmo like this when you buy the scanner.
Incidentally. the sides of the scanner are not quite perpendicular. But amazingly the sides of sheets of paper are quite often out of whack too. I think my bank actually feeds paper off a roll and trims it to size after printing, possibly to allow full bleed, or possibly because roll feed is more reliable than sheet feed.
14. Clearly you can scan a book much better if you cut the pages out of the binding first. What I did not realize initially is that it is worth taking a little extra effort to make a *straight* cut along the binding. This makes it possible to locate the pages distinctly more repeatably. If the cut is not straight, at least make sure that the scan is perpendicular to the workpiece by aligning one of the machine-cut edges of the page to one of the ruler axes, rather than trying to jam the ragged edge against the (other) ruler axis.
I used an L-shaped metal ruler to guide the knife when I cut the spine. This worked well enough. Of course, you need a cutting board under the workpiece.
15. If you don't cut the spine off, it helps a lot if you have something against the side of the scanner that the excess part of the book can rest on, so it doesn't drop out of the scanner. You will probably need such a support on both sides of the scanner.
16. The book will still tend to twist around as you close the scanner cover on it. You can try just not putting the hood down, but I think that will cause ambient-light issues. Another thing to do is use a thin, bendy metal ruler to hold the book in place while you close the lid, but you may still get some movement when you withdraw the ruler. What would be really good would be a similar device attached to the exterior of the scanner at one end so that you could just leave it in place during the scan and lift it off later.
17. Now a few miscellaneous issues.
18. How do you handle all your (paper) files *during* the scan process? As you scan you will have a perfect opportunity to mark files for disposal, which means you can't just transfer them back to where they were. I did not think this through initially. What I wound up doing is having one big bag of files which were really junk – not worth scanning. Then I had another area for files which I had already gone through and (largely) scanned, but intended eventually to discard. In this area, I kept the files in their original folders, so although they were in a big heap, they could be re-accessed if necessary. I also needed a (hopefully much smaller) area for files where it was essential to retain the paper originals – tax returns etc.
I had a fourth area for files curently being worked on – because I would realize that I could use the current setup for several sets of files, for instance.
If you are doing something like this for an office you probably should set up at least three trestle tables for this.
19. Once you think you've *finished* the scanning, how should you proceed? First you should make a *backup* of the scans! Then you need to store all the paper files you want to retain. Then go through all the items for destruction again quickly. Then destroy them. Since all trash in Cambodia is opened and examined at the curb, it must be shredded with a crosscut shredder. It may also be advisable to randomly separate each bag of shredded material into two bags, and discard the two halves at an interval of a week or so. Burning is better, but hard to manage. On the good side, Cambodians seem to burn anything anywhere, so they're unlikely to complain unless you burn down the building.
2007 May 22 [ Tue ]Google problem for people in Cambodia seems partially fixed
A few days ago I described a problem with trying to access Google from a cambodian IP: [http://www.panix.com/~dannyw/weblog/Computers/Internet/google-kh01.html]
Weirdly, the next day I noticed that the problem had been fixed. I had noticed that somebody from Google's IP range had been snooping round the site, repeatedly hitting it first from Google and then an IP in Singapore or somewhere; maybe he checked it out when he happened to notice I had a Google gripe.
The other possibility is that Google was *always* working in Windows, because it's *still* not working in Linux Firefox. I can't believe that though, because I had been noticing the problem for weeks, and I still use Windows occasionally. ...I just checked and the problem is gone from one of the machines I use most, in Windows Firefox.
Hmm. Why is it still *not* working for me in Linux Firefox? ...Hmmm... I tried turning Javascript on but it diodn't help.
2007 May 17 [ Thu ]Irritating problem with Google in Cambodia
For a year or so Google has had a Cambodian version. Apparently, as soon as it was available, the main Google page at www.google.com would notice if you accessed it from a Cambodian IP and would add the option of clicking a button for the Cambodian version.
However, something was not very well thought out: they tried to display the link in Khmer script, but on 99% of the machines I've seen it displayed as boxes. I never tracked down what the real problem was – the code for Google's main page is much more complex than its appearance suggests – but I'm guessing they tried to use Unicode, which very few browsers here are set up to support. I'm sure that if they had just tried to use Limon they would not have had the problem, but it probably doesn't fit into their setup, and it's not politically-correct. (Even if the browser happens to be set up to display the fonts correctly, the Cambodian users don't know how to enter text even if the Khmer Unicode keboard handler is installed.)
Presumably Google people noticed that almost nobody was using their link, so they decided to go the fascist route: now all attempts to reach www.google.com are redirected to www.google.com.kh. Considerately, they left a link to "Google.com in English", which is the only thing you can read on the page (for 99% of befuddled users).
Regrettably however when you click on the link (to www.google.com/ncr) it *doesn't work*; you are redirected *again* to www.google.com.kh.
I assumed this would be a temporary thing, and started accessing www.google.com by IP number. However, after a couple of weeks that IP number stopped working. Presumably somebody noticed a bunch of accesses by IP number and said "we'll fix him! we'll deny accesses by IP!". I found a different IP number which is working right now, but I fear Google may implement a system-wide fix for my "problem".
Google has an excellent reputation (although I have always maintained that they lucked into search as one of the few applications of computers where results may still be useful although provably wrong, and better yet difficult for the hoi polloi to prove wrong) but this problem makes them seem not only utterly clueless but offensively intrusive.
2007 Jan 10 [ Wed ]Debugging CGI programs
In trying to add new features to this site recently I have had to figure out, or dredge out of my fading memory, the following tricks for debugging CGI programs.
1. Of course you can run perl from the command line, like so:
perl ~dannyw/htdocs/bin/blosxom.cgi
With older versions of perl, this will automatically prompt you for CGI variables. You can enter these one after another like this:
startdate=200610
listqty=20
and finish with a ctrl-D. Except ctrl-D didn't work for me! It turns out my shell was set up to accept a ctrl-D at the shell prompt to log out, but when *not* at the shell prompt the shell just silently discarded it – not I think what the docs claimed. I turned off the logout feature and then my shell passed ctrl-D to perl as advertised.
2. You can also put the parameters in right on the command line:
perl ~dannyw/htdocs/bin/blosxom.cgi startdate=200610 listqty=20
but for some reason I could not make that work to set the "Extra path information", ie the parameter PATH_INFO which Apache passes to cgi programs in the environment and which Blosxom expects to contain the folder path eg "/Reviews/Movies".
What I had to do instead was set that parameter in my shell environment:
setenv PATH_INFO /Reviews/Movies/
which of course does not interfere with the value set by Apache if you're running from the command line – duh (it took me a long time to realize that). (The "setenv" built-in function works for my shell but probably not for yours.)
3. However, now I discovered that my suspect code did *not* bomb from the command line, so I still couldn't see the error message! So I had to put a call to the Carp module in Blosxom.cgi:
use CGI::Carp qw(fatalsToBrowser);
which told me the file I was trying to write to was "no such file". That's weird, I thought: it's there, when I check from the command line.
Then I noticed that another filespec in the same section of code – one which worked – used a different path. Aargh! In Apache's environment it doesn't understand the same filespecs that I can use in my shell – including when I run a .cgi program from the shell. I changed the filespec to match the other file's path and all was well.
It's really strange: I could have sworn that the bad filespec *did* work a few times... oh well.
2006 Dec 25 [ Mon ]Interesting vulnerability in Windows/Word
I noticed while reading the following Slashdot discussion: [http://ask.slashdot.org/askslashdot/06/12/15/0144253.shtml] that the default behavior of Windows in opening files is not what it appears to be.
You would think that Windows goes by the extension, and if it doesn't recognize the extension it asks the user to select an app. But what actually happens before that is Windows somehow checks whether it's openable by Word, and if it is Windows opens it in Word.
I just tested this by renaming a .DOC file .XXX and Word opened it without a qualm.
The significance of this is that if you are a system administrator and you have been filtering MS Office files because of the danger of macro viruses etc, you actually need to filter *any* extension.
Presumably the reason for this was to support Word Perfect users who were used to naming their files with arbitrary extensions. But surely the person who did this must have realized it was horrible for security, right? It's an example of the bad things that can happen when the company that makes the OS also makes the app.
2006 Oct 13 [ Fri ]Sorry about "how to reattach to a job"
A few days ago an article showed up by mistake on my log. I had written it years ago but found that the content wasn't correct before I posted it, so I put it aside by the expedient of changing the filedate to a date unimaginably far in the future.
That happened to be this September 27th. The blogging software obediently started displaying it.
I didn't notice for a while and when I did I wasn't sure if I should erase it or publicly disavow it as I am doing now; eventually I came down on the side of not rewriting history (for the sake of my two faithful readers – hi, Seymour and Piotr!).
The article: [http://www.panix.com/~dannyw/weblog/Computers/Opsystems/Linux/jobs01.html]
Incidentally, I still haven't really figured out the problem. I tried sending a msg to panix help, who didn't know either. However, the workaround is to use the "screen" utility, which has a built-in command to reattach to broken sessions. Do "man screen".
2006 Sep 27 [ Wed ]
How to reattach to a job
Some tasks in Unix are relatively easy: "You want to just read a text file? Sure! Just do man less!!". (You think "less? Less?? wtf?".)
Other tasks are less easy to find. In particular, certain aspects of the shell are hard to find because they are documented under the shell rather than as a separate utility. Also, even if you have a hint about what function or keyword they are related to, it is very difficult to search the gigantic man file for your shell because the name of the function tends to be short or common.
This is I guess my "mea culpa" for not finding the answer to the following problem: if you lose the connection to a login session and the server retains a suspended version of the task you were running (ie the editor, mail client, etc) how do you reattach to it?
Here's the answer, at least for the cshell. First you probably want to examine the list of suspended jobs in order to be sure which one you want:
jobs -l
Then you can restart it with fg:
fg %job
Restart a suspended vi with "%vi" or "%?string" to specify a job containing "str ing"
2006 Jul 05 [ Wed ]
Yes, you can find dual-layer DVDs in Phnom Penh -- from China
In my previous posting on DVDs available in Phnom Penh, I pointed out that all the DVD copies you see are single-layer, whatever the box says. I also speculated that genuine Chinese DVDs may be available in dual-layer.
This week I experimented. There's a large Chinese-oriented DVD/CD store on the 3rd floor of the Soraya Mall at the northeast corner; it has a lot of Chinese movies, plus software with Chinese packaging. (It even has a Korean section, which I'd never noticed anywhere else in PP.)
The staff did not speak very good English, but they had heard of the term "dual layer", although they were unclear on what it meant (as indeed I would have been up till quite recently). They started off by waving boxes with a logo saying "9" at me, until I pointed out that these did not actually say "dual layer", the way the kosher labels do, and I would not believe them if they did.
Eventually they came up with what appeared to be a kosher DVD from China *of a Chinese movie* (I was assuming that like Thailand the Chicoms protect their own industry even though they allow unfettered piracy of other countries' products, much like the USA in the 18th century). I was won over because it was labelled "Region 6" – I have never seen that on any other DVD in PP, although one sees "Region 1" quite often (and it is not true). So I crossed my fingers and hoped that the "dual layer" claim was true.
I did not ask the manager (evidently Chinese) to check it, and paid 4 USD, so I felt rather foolish until I checked it and verified that it was a dual-layer – I think the capacity was 6.5 GB. (I usually don't pay more than 2 USD for a DVD, and that day I happened to buy seven DVDs at a dollar each that were quite good.)
It's called "The Music Box": [http://www.lifeofguangzhou.com/node_10/node_35/node_109/node_266/2006/03/31/11437845481570.shtml]
There were a few more interesting elements:
1. My software actually showed it was *not* a region 6 disk; it displayed as Region 1, 2, 3... etc for apparently all regions. (It occurs to me that this may have been the *very first* disk my player has ever played which was not Region 0.)
2. Despite the extra data, and with a not very long movie, the quality was not very good. There were plenty of little specks of dirt on the image that showed it had been made from an optical print of routine theater-issue quality. I wonder if these imperfections caused a high noise floor that wasted the bitrate leaving nothing extra for the picture.
Still the image was adequate, even when zoomed in to fill the screen from the 16:9 source. Shadow detail was fair. I don't think it was simply captured from a theatrical screening.
There was one section, about two minutes long, with several read glitches. My DVD player seems to show read errors that other players don't, so I won't hold that against the DVD; however, a similar region did show up when I did a surface scan on my computer's DVD drive.
If I had to guess, I'd say it was a kosher digitization from a poor print by a service company that is inexpert. The menu system, for instance, looks more amateurish than most ripoff disks.
3. Again suggesting a theatrical print, it had Chinese and English subtitles burned onto the image; there were no separate sub tracks on the DVD. (I'm guessing the producers had US distribution in mind, and apparently there was an official DVD release in 2006-05, although I can see no reference to a theatrical run.)
Responses: 3
Name/Blog: Strider
URL: http://www.globalgentlemensclub.net/forums/index.php
Title: Admin
Comment/Excerpt: Danny Sexpat, This blog has practically no sex-related stuff on it, because a bunch of nasty little blackguards control the world and make your life miserable if you try to tell the truth about sex. Maybe the Gentlemen's Club is different, but right now I don't know. Strider
Name/Blog: The Boss
URL: http://www.panix.com/~dannyw/weblog/
Title: Loons from Global Gentlemen's Club
Comment/Excerpt: There were a lot more postings from the GGC but I deleted them -- the usual nutty child porn stuff you get from trolls. And nasty little blackguards.
Name/Blog: The Boss ggqhalbzm
URL: http://www.panix.com/~dannyw/weblog/
Title: Somebody is trying to screw up the comment system
Comment/Excerpt: Today I found about 50 junk comments. They were not spam. Not only the comments themselves, but also the URLs were clearly random generated. I've left the "excerpt" field and "blog" field from the "message" that the twerp left on this posting. lexphosce nnzuwqpr lyuvemuuc zuhejzgs []
DVD video copies in Phnom Penh
Here are some things I've noticed about the cheap video copies available in Phnom Penh. Much of this will apply to Thailand and any other country in Asia.
1. The information on the outside of the box cannot be trusted at all, even on very significant details like the subtitle or dialogue track languages. My impression is that for movies which already exist on DVD, the copy just shows a scan of the artwork from the product it was stolen from – for instance, the recent "Doom" was available with a cover with the slogan "Keiner kommt hier lebend raus!". This is then overlaid with a text box giving the language info, because the copier frequently downloads "fansubs" in other languages from the internet and adds them to the data streams.
Even when the desired languages are present on the disk, they are surprisingly often mislabelled or unlabelled. My guess is that the person who knows how to specify the language name when burning the master is not someone who *knows* a lot of languages.
2. For instance, most DVD covers show something like the "DVD 9" symbol. Even when the "logo" carries the words "dual layer", it is not. I bought a DVD writer today and asked to check it; the (very helpful) technician had suggested I bring a DVD 9 to check the dual-layer aspect, but I went to all three stores I normally frequent and could not find *one*.
At one store, the salesman said he knew of a few real DVD 9's but when I asked him to check them on a PC he discovered that only one out of 5 had more than 4.7 GB: it was 4.85 GB! I thought it was highly unlikely that anyone would burn a dual-layer DVD so close to fitting on a single-layer, and concluded it was probably a single-layer too.
Actually, when I showed up to check the drive, the tech had found what he assured me was a real dual-layer DVD. It was a Chinese movie; perhaps I should try some of the Chinese-language video stores.
3. When you buy DVDs, you should always ask the store to check the quality. I have several times discovered that the disk does not work at all in their drive; once or twice they tried several of the same title and they were all bad. My DVD player has problems with perhaps 20% of disks and cannot play them to the end (although I've been having a run of good luck lately). I therefore like to go through the chapters checking that they're all accessible, although this by no means guarantees that the disk is also usable on my drive, and I have been getting lazy on this.
Irritatingly, in one store the owner has taken it upon himself to not even demo DVDs which are not what he calls "top" quality to me. Sometimes, if I feel I will never watch the movie again, I'll put up with a cam copy, as long as the sound is reasonably clear.
4. Even the best copies have typically been compressed to fit on a single layer DVDR. My copy of "The Lion, the Witch and the Wardrobe" is mostly clear, but has a few patches with severe digitizing artifacts. Irritatingly, they do not always take the opportunity to remove the lock on the "FBI Warning" and similar infuriating stuff that you have to sit through on the legit version.
It would be nice if there were some sort of standard labelling system so that you could check the quality of a disk at a glance. Even if the *stores* didn't want to use it, the customers could theoretically *ask* for all the details, if there were some concise, standardized jargon. At the moment the only word I can use with the Cambos is "clear". One shop uses "90%" and "95%", but has not succeeded in explaning to me what those terms mean.
5. In order to help me learn Khmer, I have looked around for DVDs with Khmer audio and subtitles.
I have found *no* Hollywood movie available on DVD with Khmer subtitles. What you *can* get is VCDs with a Khmer audio track, no English, but sometimes English subtitles hardcoded on the video. I find them very hard to learn Khmer from because the Khmer seems to be a very free translation (to the extent I can figure out what they're saying at all) so the English subs are little use.
I have found a few *Khmer* movies with English and Khmer subs, but they tend to be about kings, wizards and giant snakes, so the vocabulary involved is not ideal. (Also, the quality of the English subs is not wonderful.)
I have experimented with subtitle software and it seems surprisingly easy to rip the English subtitles from a Hollywood DVD so that you have a text file which can easily be used to carry out a translation to Khmer. If I get around to it I may produce a DVD with Khmer subtitles and see how long it takes before it appears at Soriya.
When I ask vendors why DVDs are not available with Khmer subtitles they just say that Cambodia is a small market. It seems to me it would not be that difficult to set up fansubbing groups as are common in the US anime world. This could be emailed back to the manufacturers in Malaysia and Vietnam. The subtitle files are not hard to add and are amazingly small, even though they are stored on the disk as bitmaps. It's hard to see exactly how to *make money* of course, although surely Khmers would prefer DVDs with subs to no Khmer at all. Perhaps Khmers are just not very good readers and would only respond to products with Khmer audio tracks.
It occurs to me that any Khmer fansubbers would have to be very careful to wait until the kosher English subs are available from the official DVD release. Their chance of producing anything comprehensible from the subs available on zero-day releases would be zero.
6. I can see no sign of an official crackdown on copyright evasion. Around the same time as I saw reports that there was some sort of crackdown in Malaysia, I noticed a slowdown in supply, and then I saw that one store was reduced to receiving DVDs shipped in suitcases, but that's about it.
The porno isn't on display, but if you are a single man walking around without your girlfriend you will find several stores eager to demonstrate that they have it for sale (I was not brave enough to ask them to demo it however). OTOH the porno available seems pretty drab and flyspecked.
2006 May 13 [ Sat ]Weird problem with Putty pscp
"Putty" is the Windows program I use to connect to my servers using the secure ssh protocol. It's freely downloadable. It includes the "pscp" program for doing file transfers via the encrypted ssh link.
A few days ago I needed to do a file transfer but pscp failed with a strange "out of memory" error. Feeling superior, I booted my little CD with "Damn Small Linux" and ran scp, the equivalent version of pscp but otherwise completely unrelated, only to get the same error msg!
As I discovered on that same day that the internet cafe had put a virus on my USB key that was designed to capture passwords, I was a little concerned. Indeed I have not been back there since (when I told them that at least two of their machines had the virus, and they needed to install an up-to-date virus checker on their machines the guy said "Oh yeah? Uh-hmmm?") although I probably will, after a decent interval has elapsed. (Incidentally I usually add a shortcut or two to the browser on machines I use, so I can detect when they reload them: apparently not for months, which I suppose is some sort of tribute to WIndows XP/2000.)
To cut a long story short it eventually occurred to me to Google for the error message, and here's the answer: [http://the.earth.li/~sgtatham/putty/0.55/htmldoc/Chapter10.html#S10.6]
What had actually happened was that I had added a string to my Unix prompt, and bozoically added it to my .shrc instead of my .login, so that the string was generated not only on login but by every process which spawned a shell. Unknown to me, this was actually what my pscp/scp were trying to do at startup, but they interpreted the string as a call to reserve a huge amount of memory at the client end, causing the client to throw up its hands with a stupid misleading error msg. Aargh.
2006 Apr 26 [ Wed ]Another reason for my laptop overheating problem?
A few days ago my laptop abruptly died during a video game. I figured it was overheating again. As I folded it up to put it away I noticed that, as usual, the steel tabletop under it was quite hot.
Then it occurred to me that the little rubber feet originally stuck to the bottom of the case had fallen off a year or two ago. Hmm. I tried sticking some domestic rubber feet on – after cutting them to size with scissors – and the laptop now sits *much* higher above the table surface. Now you might think that reduces heat transfer to the table – lose – but the intake port for the fan is *under* the laptop, so win, big. The outlet airflow feels distinctly faster.
Now I feel a little silly for not thinking of this before. At the time I first noticed the original feet were dropping off I vaguely figured the separation was defined by some other plastic protrusions in the case molding but now I look at it they are clearly inadequate.
2006 Mar 29 [ Wed ]Surprising bug in MS Access
My girlfriend is doing a MS Access course, and asked me to help her to set up a form for an employees table. I looked at the printed version of the form, which showed a mugshot of the current employee, and looked at the instructions she had noted down from the teacher on how to set up that (image) object on the form.
"Hmm, don't really understand this." She said "oh, the teacher said don't worry about that bit."
Well, she wanted me to address her current problems – things like she didn't understand that the "retirement date" object on the form needed to be an unbound text object whose control source was the birth date field of the data table, rather than being a separate field in the table. And then I got sidetracked because the version of the date-calculating function that she had noted down from the class assumed that a year was exactly 365.25 days.
But I came back to that image field. There was definitely something funny about it. It turns out that this is a well-known issue: [http://www.jamiessoftware.tk/articles/handlingimages.html]
The above link is quite good but does not give you the full picture. In particular it does not apologize for the misinformation in Access's own help files. It also does not give you a full sense of how many webpages you can find relating to this which give misinformation or lack the full story.
Microsoft has several pages relating to this issue. This one is probably the most helpful: [http://support.microsoft.com/kb/q210100/#EJADAAA]
Here is my summation of what I discovered:
1. Access allows you to put image files into the database as an OLE object, but it seems to actually store them as uncompressed .BMPs. When I started experimenting with this I was using fullsize 2 MP .jpgs. I have since stopped using the OLE object technique, but my tiny database file – with only six records – is still over 20 MB. I suppose I need to compact it somehow, but I haven't checked that out yet. So as the first link suggests, it's really impractical. (Incidentally, it occurs to me that even a few *static* images could really bloat up a database.)
2. Even when I used the OLE object technique, I *still* wasn't able to display images on the form (except static images of course). I don't know why. Some references on the web say that you need to install Microsoft Picture Editor to make sure Access can find an OLE server for your graphics file, but my system already has umpteen apps registered to handle .jpgs. I tried changing the registered app to various things including IE, and it still wouldn't work.
3. Anyway, I wanted to specify the image file by storing the path and filename in a text field. It turns out that you need to put code in both the form itself and a module in the database. (As described in the Microsoft link above.) This was not hard to do by cutting-and-pasting, but the fieldnames and tablenames are hard-coded for the example, so you would need to duplicate it for every image field in your database. (I think – actually, I really don't understand what the code is doing.)
4. Even then, it only worked if your filespec included the *full* path to the file – eg something like "L:\data\msaccess\imgtest\pics\danny-small.jpg" This contradicted the help file, which specifically suggested that a path relative to the location of the database file would work. Eg, if the database file was "L:\data\msaccess\imgtest\test01.mdb" then your filespec could be "pics\danny-small.jpg". I thought this really sucked – it would make it very hard to move the database file around. Eventually, I realized that the following *does* work: "imgtest\pics\danny-small.jpg"
5. I almost forgot the first way I made it work. The MS documentation refers to "data access pages" and for a long time I assumed this was some sort of generic term for "forms". However, in microsoftspeak it refers to a specific kind of database object which you can create. It is like a form, but can be accessed by Internet Explorer. I don't know exactly how it's supposed to work if you have multiple clients; presumably the form actually tells IE to use the client's locally installed copy of Access to access (aargh) the Access file on a simple fileserver. Anyhow, it works for displaying images, with no need to add code.
6. I don't remember the details, but Access did not seem to automatically resize an image to fit the control size you set on the form. I think maybe the docs are wrong. I needed to resample all of my images to make sure they fit in the pixel area allocated on the form.
7. The "Northwind" sample database somehow manages to make images work without using the sample code. On the other hand, they are very low-resolution images. Additionally, they are only used to illustrate product *categories*, which if you think about it is pointless – the database does *not* have images of the products themselves, probably because then it would need to be about 500 MB.
All in all, I thought this behavior was incredibly buggy and kludgy. I would have thought that handling images was not rocket science.
Also, it makes me wonder whether *all* the documentation produced for software is produced based on its *specs* and never tested against actual behavior.
Also, I would like to point out that Microsoft appears to have a policy of taking over a common term whether it has a generic or specific meaning and redefining it. For instance, "Word". More interestingly "domain", which has quite different senses on the Internet and in connection with Microsoft server setup, even though (obviously) you are usually discussing situations where either sense might be intended.
Likewise, here, "data access pages". It makes me feel less irritated about Unix naming conventions. When you encounter a name like "snort" you don't know what it means, but you *do* know it's a name and you need to look it up.
Finally, I suspect that my girlfriend's teacher never got images to work on his form at all.
2006 Mar 02 [ Thu ]Fun reinstalling Windows 2000
I ran out of space on my hard drive and figured I should reinstall Windows – maybe that would fix the problem with the scanner.
I ran into several problems though.
One big one appears to be overheating. (I say "appears to be" because the error pattern in the LEDs on this machine never matches the manual.) Presumably during the install the setup software doesn't handle the powersaving stuff, runs the CPU at max, and causes overheating within a few minutes. Still, I've managed to do it before... but now I come to think of it, that was probably in an airconditioned room.
Still, the *irritating* thing was being unable to copy the files from the setup folder on the CD, boot and then just run everything from hd. I was stymied in several ways:
1. I had put the setup files on the NTFS partition. But my "Ultimate Boot CD" needs to load a memory manager to load NTFS support... and it crashed on my laptop hardware.
2. I tried copying the files to a DOS partition. This would have worked, if Windows 2000 Setup could run under DOS. D'oh!
3. I tried running Bart PE, which could see and write to the NTFS partition just fine... but Windows 2000 Setup refused to run an install on a machine running a later version of Windows. D'oh! (If I remember rightly Bart PE has to be used with XP – you can't make it boot W2K.)
Oh well. Back to the Windows CD and SP4... overheats again... aargh...
2006 Feb 15 [ Wed ]Firefox memory problem a "feature"?
According to this Slashdot thread: [http://developers.slashdot.org/article.pl?sid=06/02/14/2154224] Firefox takes more and more memory because it keeps old pages in RAM (so that they can be reloaded faster) even after you close their tabs.
I had been getting problems using Damn Small Linux because the Firefox was using more and more memory – not good when you can't page to disk.
I'm running IE now (aargh) so I can't check it, but apparently you can fix this by putting about:config in the URL box, then scrolling to this value:
browser.sessionhistory.max_total_viewers
and set the value to 0.
...hmmm. Later posts in the Slashdot thread deny it. Oh well, I'll try the above fix anyway and see what happens.
2006 Feb 14 [ Tue ]The "Honeynet" project has various anti-hacking tools
Their main page is notr very exciting. Try this one instead: [http://www.honeynet.org/tools/index.html]
A "honeynet" is a system of servers which are set up to act as "honeypots" – ie, machines which will appear to be hackable so that hackers will find them and attempt to hack into them. But they have special software running which allows a great deal of information to be gathered *from the hacker*. For instance, as soon as a hacker gets into a machine, he will download his own tools package somehow. Often this process will allow the honeynet machine to capture the machine name, user name and password of the accounts that he is using, which provides both a lead to his identity and a way to gather more information (eg by logging into his IRC group).
honeynet.org provides a CD which can be used to easily set up a honeynet server. They also provide the tools individually, as in the link at the top.
It strikes me that the dividing line between hacker tools and anti-hacker tools is rather small. For instance, if a hacker manages to subvert DNS (or ARP), client workstations can be fooled into thinking the hacker's machine is their normal server. The hacker merely has to use the honeynet software on his machine: when they attempt to log in, their usernames and passwords can be captured.
One of the individual tools is "sebek": [http://www.honeynet.org/tools/sebek/]
It has two components. The first is a client that runs on the honeypots, its purpose is to capture all of the attackers activities (keystrokes, file uploads, passwords) then covertly send the data to the server. The second component is the server which collects the data from the honeypots.
If this software can be used to gather all this info from relatively wily hackers, it can certainly be used against normal users. Like your girlfriend. Hmmm.
2006 Jan 31 [ Tue ]Microsoft is apparently still allowing sites to share your cookies
I have posted about this before: [http://www.panix.com/~dannyw/weblog/Computers/Internet/cookieexploit02.html]
A description on c/net: [http://news.com.com/2100-1023-245680.html]
A recent posting on Slashdot says that this is still happening: [http://it.slashdot.org/comments.pl?sid=175615&cid=14603033]
The existence of this security issue makes a mockery of Microsoft's public pledges to concentrate on security – as do many other similar issues.
2006 Jan 06 [ Fri ]My laptop woes abate
My Atec laptop has had various problems, although it's never needed an expensive repair.
Recently it started shutting down very rapidly, amid apparently bogus "low-battery" warnings. In fear and trembling, I disassembled it. I was expecting to find the fan compartment was clogged with dust, but it was almost clean. However, there was almost no heatsink compound detectable and what was visible had hardened. I figured this matched up with the fact that the case of the machine got red-hot but the exhaust air was scarcely warm, so I went looking for heatsink compound.
I had assumed this would be available because there are a ton of places that assemble PCs, but I had no luck. Eventually a tech at PTC on Monivong took pity on me and gave me a little hypodermic for free (hmmm... I wonder if he filled it with toothpaste... nah...).
I opened up the laptop again and reassembled it with the heatsink compound. (I have at least a dozen screws left over – no joke.)
Then I didn't get the fast bogus low-battery shutdowns; now I was getting *instantaneous* shutdowns.
I figured that when I reassembled the case I had left something loose enough that a slight flexing could short out the power supply. I experimented with using an external mouse and keybd, and that worked very well.
After a while I tried running without the external devices, and the laptop *still* worked. Then it occurred to me that the period when I suffered the instananeous shutdowns was exceptionally cold and dry, with nighttime temperatures going as low as 70 F; and right around the start of the problem I had actually felt a static zap in an air-conditioned store. Hmmm! I guess what I had actually done was loosened one of the earthing straps which, when you open up a laptop, you can always see were added in desperation at the last moment before the ship date, and get in your way until the tacked-on solder joint snaps. Static zap: foom. I guess I was lucky it just reset the machine rather than burning something out.
Anyhow, now it seems to be working fine... touch wood. I'm hoping the earthing wire got jiggled back in place; but it's probably just hot and humid now.
2005 Dec 26 [ Mon ]Actually, Joomla *can* work with Moodle
In a previous article I confessed that I had mistakenly assumed that Moodle was module-compatible with Joomla: [http://www.panix.com/~dannyw/weblog/Computers/Internet/Joomla]
Actually Moodle *does* have a degree of compatibility with Joomla, in that Moodle can be installed so that it runs "inside" Joomla, at least as far as the user sees it. Right now that capability seems to be poorly documented and liable to break things, but apparently some people have made it work. For instance, they can share the user/password database. (phpbb can do the same thing, although regrettably phpbb has fewer restrictions on usernames so you usually can't import existing phpbb users into the Joomla system.)
Actually, I think it's probably better *not* to share the user database for my purposes, but it's good to know it's not impossible.
See eg discussion here: [http://moodle.org/mod/forum/discuss.php?d=20080]
Apparently features like this are called "integrations" in Moodlespeak: [http://download.moodle.org/modules/integrations.php]
I had been vaguely looking for this stuff under "modules". Silly me!
The "filters" category is a little thin, but you should check it out. For instance, you can install a filter that turns TeX code into GIF images: hmmm...
2005 Dec 20 [ Tue ]I have suddenly realized why my Corel Draw pics were off by a pixel
A few weeks ago I was creating some images that needed to be precisely scaled, so I told Corel Draw they were 4 ins by 6 ins at 300 dpi. Then I exported to jpeg. Fume! They were 1801 by 1201 pixels! I could see there was an off-by-one error somewhere (duh) but I couldn't see where it was coming in.
Yesterday I was working on a similar project, but this time I had created some patterns which I positioned outside the nominal page size, in part. To my surprise, when I exported to jpeg, the borders of the image were the edges of the patterns, *not* the nominal page size.
In other words – which I had not suspected – Corel Draw does *not* crop your image automatically to its nominal size.
Musing about that today I suddenly saw the relevance to my earlier problem. I had set the guidelines to 0 and 6 ins, horizontally. But that makes the righthand edge of the image (when you snap to it, which is the reason you have guidelines) set to 6.000 ins. But that's actually just *over* six ins, when you think about it (assuming that 0.000 is the lefthand edge of the first inch, ie your six inches are inch 0, 1, 2, 3, 4 and 5). Aargh! So now I have to carefully set my guidelines to 6.000 - 1/300 ins! I'm not even sure I *can* set that accurately (although I think I saw some global setting for x/y resolution somewhere that might make it doable).
Incidentally, the off-by-one doesn't *really, really* matter because the processor is actually printing to a true photo-chemistry printer anyway, so they automatically scale up the image by a couple of per cent to make sure the projected image fills the borders of the print paper. So the scaling is off by some small but irritating amount – I think about 3 per cent – anyway. Oh well. At least they don't decide they can crop the shot better than I can.
2005 Dec 14 [ Wed ]I made a silly mistake with Joomla
No, not with the patch (as far as I know just yet).
For some reason I got it into my head that Joomla was compatible not just with Mambo but *also* Moodle. I'm not good with names so I guess I just remembered them as "silly sounding name N". As far as I can see, while Joomla and Mambo are still compatible (at least for now), Moodle is a *completely separate product*.
This is a pity because actually *Moodle* has the education features I need.
Fortunately Moodle was easy to install and has been flawless so far (not very far) and I think I may maintain *both* systems for a while. It might be good to have a *total* separation between the public features and the features for paying customers.
Patching Joomla 1.03 to 1.04
I was puzzled about how to *apply* this patch, even though pretty clearly it was the file I wanted (the filename was after all Joomla_1.0.3_to_1.0.4-Stable-Patch_Package.tar.gz). It just said something about "put this in your Joomla directory".
Having dealt with tar files before, I suspiciously checked the directory structure inside the tar file. It corresponded to the directory structure inside the Joomla directory. So (in fear and trembling) I went ahead and ran tar -xvzf | less.
After maing the disappointing discovery that on my new server "less" appears to be symlinked to "more" instead of the other way round, I could see no error messages.
Then when I tried to run it from the browser, everything seems to work fine and it does say version 1.04.
I don't know quite how this really worked. For instance, I noticed a new htaccess file in there, but I assume it has to be manually copied to .htaccess.
Likewise, the files are marked owner/group "pasamio" inside the tar file but of ocurse were created with my default owner/group when I untarred them. I think I'm just lucky that happens to be OK for this webserver.
2005 Dec 10 [ Sat ]Setting up Joomla
I recently set up another site. One reason I chose my hosting vendor was that they had Joomla via web-based install: I figured I didn't really need web-based (in fact I hate web interfaces for admin) but it probably meant that they had several users already running it.
It basically worked ok. Some observations so far:
1. Joomla actually has surprisingly little documentation. For instance, I couldn't find anything that explained which visible items were part of the base install and which were part of the template.
2. The site logo, at least with one template I checked, isn't even defined by HTML, but in css. It took me a long while to find it (it didn't help that I kept thinking there must be an easier way to track it down as I pored through HTML, PHP and CSS).
3. Even when features *are* accessible via the administrator's web interface, it's not obvious *where* they are in the interface. I found myself wishing for some sort of feature where you could hover the mouse pointer over an element and it would pop up a little window that told you how to edit it. Or documentation.
4. So far it's been stable. I'm fairly impressed.
5. The update process is a little scary. Why do they give you a windowing interface and handholding to do trivial stuff you could do faster at the command line, but to do a security update you have to manually overwrite every changed file on the site? Oh and by the way, there's no readme file to explain exactly how to overwrite it. What happens if you *edited* some of those files?
6. I'm surprised – considering all the hype about Joomla – how few features there are. I was assuming there would be software to create class packages; payment modules; test result analysis... but about all there is seems to be timed release of content and user registration.
7. Huh – I noticed something in the sourcecode about creating PDFs, and now I found that if a user goes below a certain level in the hierarchy, a content item gets a clickable PDF icon – and it seems to work (although for some reason Firefox crashes if Adobe Reader is *already open* – go figure).
2005 Nov 06 [ Sun ]How to repeat a long filespec in a command line
I have been looking for how to do this for *years*. You often have to change the name of a file in another directory. That's easy in Windows, but in Unix you have to key in the whole blasted filespec again (and hope you don't make a typo): mv /home/d/dannyw/blosxom/data/images/phnompenhtrip/crossing1.jpg... aaagh, I can't go on.
In a Linux journal article on the *bash* shell, I found this:
[http://www.linuxjournal.com/article/7385]
echo {one,two,three}fish
onefish twofish threefish
echo /etc/httpd/conf/httpd.conf{,.bak}
/etc/httpd/conf/httpd.conf /etc/httpd/conf/httpd.conf.bak
Those are *wiggly* brackets, and they create *multiple* parameters to your command line, so you can do something like:
mv /etc/httpd/conf/httpd.conf{,.bak}
mv /etc/httpd/conf/httpd.conf /etc/httpd/conf/httpd.conf.bak
I tried it under tcsh, and it still worked.
Perfecto! If only I can remember it!
Copyright © 2003-2010 Alternate Worlds Publishing, Boston MA USA
The Little Trouble Girl font is copyright Blue Vinyl Fonts vendorurl: www.bvfonts.com
