The web pages and images in the examples below were produced by LATEX2HTML-FU from the latex source, with absolutely no manual editing of the produced web pages or their images.
However, I used the wiz
’s headers and footers option to
put my name to the left in the top bar and the FAMU and FSU logos to
the right in the bottom bar. I swiped these logos from the college
web site, but unfortunately, they were made against a white
background, not blue like the default bottom bar. So, I used the
color scheme option of the wiz to change the color of the bottom bar
to white. This did however not give any visual separation between
the bottom navigation buttons and the logos, so I used the headers
and footers option to append the html code <BR><HR>
(line
break and horizontal ruler) to the bottom left and right button
rows. (In Firefox, you can see that the two rulers miss each other
by a pixel. Maybe, someday I will create a separate template.html
with a single ruler.) For the rest, colors, font sizes, icons,
etcetera are all LATEX2HTML-FU defaults.
I process this document from Unix (where I keep it) using teTeX and
makel2h
, which I personally continue to prefer over
l2h
: it is less of a mess if the web pages are in a
separate folder. Processing takes a couple of minutes.
I also made a colorized version of this document using the wiz. (I am indepted to the Anry Color Picker for suggesting a more pleasing color scheme than the blue/yellow I had initially come up with.)
This is a larger document, with at the time of writing well over 1,000 printed pages, 7000 different math images, 280 figures, 10 or so animations (using rawhtml inside the tex document,) a table of contents, an index, lists of figures and tables. All not including the hyperlinked solution manual.
The main problem I encountered in this document was that in the original document, I had put some figures as “displaymath” to prevent LATEX from moving them. (The “h” option of the LATEX figure environment does not mean put the figure “here”. It means, please, please, put the figure here if you really cannot find a place that you think better of.) Unfortunately, the figure quality produced by LATEX2HTML was completely unacceptable to me due to the failure of Ghostview to properly anti-alias embedded bitmaps. So I had to use the extrascale trick described in subsubsection 9.20.20, but that only works in figure environments, not displaymath ones. So LATEX again started putting my figures where they should not be, (in the pdf version of the document,) requiring manual insertion of page breaks in the LATEX source. Eventually I discovered the “H” option of the float package to fix this. You learn something new about LATEX every day.
Versions of this document have been created using teTeX on Unix, using emTeX on an Windows XP machine, and using MiKTeX 2.5 final, with hyperref turned off as described in subsubsection 9.20.12, on another Windows XP machine. Earlier MiKTeX version (2.5 beta and 2.4), fpTeX, and XEmTeX were also successfully tested on this document. Initial version were produced on Windows 98, but I do not do that anymore because of the speed difference. Processing from scratch on my single processor Windows XP portable takes about 20 minutes, almost all of which goes into creating the 2200 images. Processing on the Unix cluster takes considerable longer, maybe twice, probably because the many thousands of files involved must be transferred over the network. Processing minor changes that do not require many images to be remade takes only a few minutes.
The wiz was used to set the font size to 19px, (LATEX document font size was 12pt), to swap the textual “next” link to the right side of the page, and to set the web page splitting level to -2. For the colorized versions, the wiz was also used to change the colors, generating the recolored navigation buttons automatically from the blue ones. The background was set to FFF0A6, the bar colors to FF5500, and the title colors to 660000. The text in the bars was separately set, using the individual properties item of the wiz, to FFF8D2, since I thought the default black too unimpressive.
Since I wanted to add stuff to the format selection screen, (such as
links to the colorized version,) I copied over select.html
from the files
subfolder of the l2hsup
folder to
the index.tex
folder, renaming it select0.html
,
and made the additions to it manually.
No document segmenting was used in this or any other example; I find that even a large document like this processes perfectly fine from a single LATEX source on all machines I tried. However, an 1999 version of LATEX crashed on the size of the document.
This was the thesis from hell, LATEX2HTML-wise. Used a
proprietary “FSU_DISS” style that widely deviates
from standard LATEX conventions. The figures have been pushed
together by trial and error from separate .eps fragments, big and
small, using a mixture of \special
, \vspace
, and
\hspace
commands, and paragraph breaks, all in a document
with a nonstandard line spacing. It had two abstracts, created
using some special \abstract
command. It included title
pages, signature pages, and other required forms.
Converting “as is” was hopeless, so I converted it to
the standard LATEX “book” class for the web
version. That was a very quick process, but unfortunately, the book
format has no abstract. So I ended up turning the (single) abstract
into a \chapter*
. Not elegant, but it seemed the best I
could do.
I also put the contents, and lists of figures and tables at the end instead of at the start, where I found them distracting in the web pages.
Another problem was that the “FSU_DISS” style uses different line-spacings and fonts than the book class, (and I did not know which ones). So, with the figures pushed together from fragments using paragraph breaks, various labels were not properly centered on their graphics in the book version. Also, some bounding boxes did not include all of the figure. One was even completely off the page. This could be fixed by using the no-EPS version of LATEX2HTML, but that still left the labels in the wrong place. So, I ended up turning 14 of the 70 figures into proper single .eps files with the Ghostview procedure described in subsubsection 9.20.11 (using the FSU_DISS version of the document that had the right linebreaks.)
One figure environment had two figures with separate captions. I had to split it into two figure environments to keep LATEX2HTML from including the first caption in the image. (I believe another way to handle these things may be through the htmlimage command, but I did not try.)
Continued figures had the text “continued” in their
figure environment. I had to put this into a caption to prevent it
from ending up as part of the image, and use a
\addtocounter{figure}{-1}
in front of the figure
environment to keep the figure number from increasing.
I made the navigation icons manually with MS Paint, using the
Monotype Corsiva font, and converted these bmps to gifs using the
bmptogif script found in the l2hsup
folder. I then swapped
them in using the wiz. Feel free to swipe them off the web page and
swap them in with the wiz if you like them. I also used the wiz to
put the navigation icons at one side of the page, and the textual
“next” link to the other side, which I think works
pretty well. A one point border was added around the bars to make
them look less flat; the procedure is in subsubsection
9.24. The font size was set using the wiz to 21px. I
overspecified the document font size as 13.2pt (true value in the
LATEX source was 12pt,) since the math seemed a little bit bigger
than the normal text to me. (Maybe I shouldn't have? The point is
that math capitals are disproporionally big compared to math lower
case, you cannot get both to match the browser font size. Overall,
I think the true value, as in the quantum text above, may be
preferable.)
In August, 2006, I used the newly enhanced jpeg support to replace the previous small gif figure thumbnails into full size jpegs, at a size that you can still download over a phone line. I put a
\htmlimage{thumbnail=1,extrascale=3}{}or
\htmlimage{thumbnail=1}{}command in each figure environment, depending on whether they contained a bitmap image like the color figures or just vector graphics like the plots. Then I used the wiz to set the figure scale factor to 2. The net effect is that the figures appear in the document as full size thumbnails, and if you click them, you get a double size image to study more closely.
I got rid of the old gif figures using clrl2h all
, and
set the jpeg quality to a high value using
set L2H_JPGQ=85since I did not want to skimp on quality in someone else's thesis. (Unix users use
setenv
instead of set
and a space
instead of an equals sign.) The jpegs are still much smaller than
the corresponding gifs; some of the streamline gifs topped out at
almost a megabyte! Finally, I processed with makel2h
.
Note that this document is not a good example of how LATEX2HTML-FU normally handles math, (unlike the quantum mechanics text and the Ph.D. thesis above.) Some math is build out of text and images, not just images, by loading the math extension. Supposedly, if you have the Century Schoolbook font installed, the math text and image symbols will be consistent, and download time will likely be reduced. LATEX2HTML-FU assumes that few browsers have that font, so by default it does not load the math extension. Also, some math looks fuzzy since it is oversized to allow higher quality printing. LATEX2HTML-FU does not by default do that, it assumes that you will provide a pdf version for high quality printing instead. LATEX2HTML-FU normally does a better job at aligning eqnarrays, but the math extension disables that, as well as some white space around the formulae.
I also created a grey scale version. In this version, I used the wiz to swap back in the 3D icons that come with the original LATEX2HTML and to set the background color to the 75% grey page color for which these icons were made. I do think the grey backgound makes the document pleasurable to read on screen; I suspect it reduces eye strain. The title and bar colors were suggested by the Anry Color Picker.
I did take some cosmetic liberties with the source document here.
In particular, I used an extrascale
in the
htmlimage
command to anti-alias the embedded image. See
subsubsection 9.20.20 for more details. If you compare
this figure with
the version above, you
will see that it does make a difference. Also, normally
LATEX2HTML-FU would expand the caption of the figure to at least
75% of the page width, but since the figure has been given a
border, it cannot do that. (Browsers would expand the border along
with the caption.) So I moved the figure to the side of the web
page, again using the htmlimage command, to avoid all the empty
blank space around it. (It would have been even better if the
figure would not have ended up next to nonwrappable verbatim text,
but I could not help that.) I also replaced the pink and purple
balls in the itemized lists with other colors since they clashed
badly with the grey background. I find that
white balls work quite well. And
I updated the bibliography images.
I included links to the setup files that are in the index.tex folder to get you started if you like this format. However, you can easily create them yourself using the wiz and change whatever you like. Specify the background color as BFBFBF, the title color as 660000, the ruler colors as 3F3F3F, the bar backgrounds as 3F3F3F and foregrounds as BFBFBF, including the bar link colors. Add a four point border around the bars as described in subsubsection 9.24.
In addition, I recreated the frames version. This allowed me to test the frames generation mechanism (and fix a few bugs there too.) For example, you can click on the index button and easily examine, say, the parts where figures are discussed. Also note the footnotes frame. See subsubsection 9.26 if you want to put your own documents as frames like this.
In processing the original segmented document, the provided
makefiles did not work even under Unix, for which they were
supposedly designed. I made some attempts to try to figure them
out, but just got hopelessly confused. So I simply combined all
document fragments together into a standard single LATEX source
and processed that normally using makel2h
. Needless to
say, this 110 page document too processes just fine without
segmenting.
These case studies show that LATEX2HTML-FU can reliably process proper LATEX documents into professionally looking, massively hyperlinked web pages. It is however not uncommon that some adjusting of the LATEX document is needed, especially for nonstandard LATEX usage.