Infinite loop in pgfplots when starting and ending values of domain are equal

[dlglin]

The following MWE leads to LaTeX running away when generating a hardcopy.

DOCUMENT();

loadMacros('PGstandard.pl', 'PGML.pl', 'plots.pl', 'PGcourse.pl');

$plot = Plot();

$plot->add_function('x', 'x', 2, 2);

BEGIN_PGML
>>[!!]{$plot}{400}<<
END_PGML

ENDDOCUMENT();

I have attached a copy of the .tex file that this generates to produce the graph. LaTeX reports "TeX capacity exceeded" on line 76. It looks like the domain=2:2 option is causing an infinite loop.

This appears to be a bug in pgfplots if the domain starts and ends at the same point. Should we be protecting against this?

A related note: both JSXGraph and pgfplots are okay with the domain being expressed in the wrong order (i.e. $plot->add_function('x','x',2,1); draws the graph from 1 to 2. Is there any reason to try and prevent this?

BTW this arose as I was updating code from an old OPL problem (Library/ASU-topics/setLimitConcepts/3-2-36.pg) where they used this construct to draw a solid circle by plotting a function that started and ended at the same point.

pgfplots.txt

[drgrice1]

When I test this, the hardcopy fails after a few seconds, but it doesn't cause any serious issues. If I run xelatex or pdflatex on the image.tex file generated by the LaTeXImage.pm module, the process dies in a few seconds with the error TeX capacity exceeded. For you does the xelatex or pdflatex process not terminate?

This is not the only way that this can occur. There are lots of ways using TikZ to exceed the TeX capacity. I don't really think that there is anything that can be done about that. Problem authors just need to debug their TeX code.

[somiaj]

Though this has come up before. We have talked about seeing if it was possible to add a time out to keep pdflatex/xelatex from getting in an infinite loop. In this case unsure if we should check for every possible issue that may cause TikZ to exceed TeX capacity, but having a general solution that will kill the xelatex/pdflatex process could help prevent server load getting caught up due to a bad problem. The time out should be a function of the number of problems being generated (to deal with creating hard copy of multiple sets/users).

[drgrice1]

I really don't think that checking for issues that cause TikZ to exceed TeX capacity is really even generally possible.

I also don't see hung xelatex/pdflatex processes anymore. So I am not sure how that happens, or if it does anymore.

[dlglin]

The LaTeX process did terminate as soon as it exceeded capacity. I'm working on an old machine so this took about 15sec. Until it terminated it had the processor pegged at 100%, so I'd be worried if this happened in a production environment with a large class.

I agree that we can't (and shouldn't) analyze TikZ code for things that will have trouble compiling, but plots.pl is generating its own LaTeX code based on its syntax, so some effort can be made to make sure that this code compiles. In this specific circumstance some code that works in pgGraphMacros.pl doesn't work in plots.pl, so techniques that were used in old problems would need to be reworked.

One thing that I would like to see in the POD and/or sample problems for plots.pl would be an example of how to plot an individual point.