Pretty silly stuff; being a bit distressed by all the overfull hboxes (lines that are too long, where hyphenation is hard) that appeared as soon as a small change was made to textwidth of a several-hundred page document, I was vaguely interested to see the variation with textwidth of some hbox statistics, for several typefaces.
The main question: one expects a general trend of better line-fitting with longer lines (more flexibility in finding space) but if words tend to have certain more common lengths, perhaps there will be some `local minima'. Are there?
Also, what difference do changes of typeface, e.g. palatino, newcentury etc. have? Some make significant differences to size, so can't be directly compared. Let's show them anyway.
Here are the scripts and data used for the plots. The shell script varies textwidth (set in this case in a class file) and typeface (set in the .tex file), then runs latex, recording all the numbers for overfull hboxes. The .m script (octave) reads the output files and plots the data.
Plots of results follow. Only integer values have been used in the variation of textwidth millimetres. The results with \usepackage{microtype} were the same as without (surprising? -- should try re-doing this!).
Conclusion (for the one, non-mathematical, picture-free document that
I studied...):
CMR (the `standard' Computer Modern of Latex) gives the fewest
overfull hboxes, followed closely by Times. Palatino, and worse still
NewCentury, have several times the number of overfull hboxes, even at
the large textwidths.
Final conclusion: leave it CMR for me, as I like that; leave it on
Times for the elderly recipient, as that seems rather heavier type
yet won't cause me to spend all day hinting at hyphenating the
last two letters off a short word...
Page started: 2008-11-24
Last change: 2008-11-24