[OS X TeX] Re: The microtype mystery

[Josep Maria Font]

El 20/08/2010, a las 17:26, Herbert Schulz escribió:

> On Aug 20, 2010, at 10:02 AM, George Gratzer wrote:
> 
>> I just woke up that my problem is solved!
>> 
>>> Interestingly, the amssymb package---which loads the amsfonts package---appears to have all the symbols that latexsym contains so there is no need for the duplication and the speed difference is obvious. Am I wrong here?
>> 
>> First, Herb you are wrong. My recollection is that latexsym defines about a dozen symbols
>> that amsfonts does not. It seems, however, that I do not need any one of them. So I commented out
> 
> I'm only going by what Kopka & Daly have in `Guide to LaTeX'. Also, the only difference between amssymb and latexsym noted in `The Comprehensive LaTeX Symbol List' is that a different glyph is used for \Diamond in the two packages; amsfonts (read by the amssymb package on loading) uses the same glyph and \lozenge.

This is precisely *my* reason to load latexsym (*after* amssymb): I want a real Diamond and not a lozenge; they *are* different things! I have always considered this a *bug* in amssymb! \Box and \Diamond are two extremely common symbols in modal logic... (btw, I redefine \Box to \square which, although slightly bigger, gets aligned with the baseline...).

Now, Martin, thank you very much for having discovered this issue. Here are my timings with the iMac, for the 16-page paper:

Without microtype (but with latexsym):
real	0m1.035s
user	0m1.002s
sys	0m0.026s

With microtype and NO latexsym:
real	0m5.538s
user	0m5.504s
sys	0m0.029s

With microtype AND latexsym:
real	2m16.795s
user	2m16.659s
sys	0m0.160s

I would say here the difference, a factor of 4, is more or less to be expected.

However, for the 133-page paper: 

Without microtype (but with latexsym): 
real	0m4.266s
user	0m2.422s
sys	0m0.058s

With microtype but NO latexsym:
real	1m30.156s
user	1m28.743s
sys	0m0.069s

With microtype AND latexsym:
real	3m39.757s
user	3m37.741s
sys	0m0.151s

Here the difference between first and second runs is around 22 times, this seems still too much, doesn't it ? For other interested people, the used packages in this case are (many intermediate lines deleted):

\documentclass[a4paper]{article}
\usepackage{fixltx2e}
\usepackage[latin1]{inputenc}            
\usepackage[english]{babel}              
\usepackage{amsmath}
\usepackage{amssymb}   
\usepackage{bm}                          
\usepackage{enumerate}
\usepackage{verbatim}                    
\usepackage{mathtools}
\usepackage{url}
\usepackage[pdftex,bookmarks,bookmarksnumbered,linktocpage,
         colorlinks,linkcolor=blue,citecolor=blue,pagebackref]{hyperref}
\usepackage{microtype}                 
\usepackage[slantedGreek]{mathpazo}      
\usepackage[scaled]{helvet}              
\usepackage{courier}                     
\usepackage[T1]{fontenc}  
\usepackage{textcomp}                    
\usepackage[myheadings]{fullpage}
\usepackage{fancyhdr}
\usepackage[amsmath,thmmarks]{ntheorem} 

Another package I suspect is {bm}, which I think constructs its symbols on the fly, unless one uses \bmdefine; I use it in many cases, but perhaps there remain several isolated \bm commands here and there...

Anyway, now the figures are closer to accceptable, especially if I \include single chapters. For me, it remains to find a workable \Diamond definition...

Best,


JMaF