The trap for L^ATEX convertors

Sergej V. Znamenskij

svz@latex.pereslavl.ru

This file contains absolutely legal and ordinary (thought somevhere miningless) mathematical constructions in L^ATEX text to facilitate visual detection of image alignement algorithm flaws in L^ATEX to HTML convertion.

Lines alignement problem

The problem comes with the large height and (or) depth formulae. Usually such a formula as $x_s^{\int\limits_a^{N^1}\, dx}$ for example has larger height, then depth. As a corollary, converto may add inproper large gap (vskip) before the next line. The case of equal height and width $\left(x_s^{\int\limits_a^{N^1}\, dx}\right)$ is more controllable, as far as we do not try to get images scaled togater with text changing. The other most rare and very hard to resolve situation is a formula with a large depth, but a very small height: $y=1-\dfrac{\alpha}{1-\dfrac{2}{3-\dfrac{4}{5-x}}}$. I will appreciate any idea to keep proper interline distance in such a case without additional vskip over this formula thought suppoce it to be impossible: I failed to find a hint in html specifications how to force browser to leave a proper place lower image witout extra gup over it.

Inline alignement problems

All dots in the line are to be exactly alighned for any browser and selected text size.

. $.\sum\limits^n1.$. $.\prod\limits_n2.$. $.\dfrac{\alpha\otimes\mu}{\gamma}.$.$.1.$. $.\dfrac {\rightharpoonup}{A\otimes B}.$.

The simple convertion algorithm can fail due to the following reasons:

• HTML lacks opportunity to reproduce L^ATEX layout directly;
• Some kind of HTML markup has different visual representation in different browser versions;
• Usable image convertion libraries and tools do not care reference point position in the image while cropping;
• L^ATEX box dimensions (width, height and depth) usually differs from its visual width, height, or depth.

See the source.