(a) Getting the marking scale without latex compiling does not seem possible to me, since shuffling, macro expansions and so on are done by latex.

(b) This could work for questions which are not random (shuffling the answers is OK, but not questions with random numerical values for example). However this can lead to errors, since you will get a marking scale different from the source file's one, and when you update the marking scale with a latex run, all changes will be lost.

(b) I see what you mean. If one knows what they are doing, they can change the TeX file and the database at the same time. I was able to do this when I realized that I had written a Question ID with a typo in it.

(a) I get it. Is the marking scale constructed by reading amc-compiled.pdf? Or one of the other files like amc-compiled.amc? If it's the latter, maybe there's a way to have TeX expand the macros the document (writing to the .amc file in the process), but not write the PDF. I'll look into that.

There are other workarounds I can see. Our process was to scan the multiple choice pages all at once, while grading the free response/open questions one page at a time. I could make two exams, one for multiple choice and one for free response. The multiple choice exam could have une douzaine or so copies. The free response exam would still need unique copies, but since it would have fewer pages and less text, it would be faster to generate.

If one knows what they are doing, they can change the TeX file and the database at the same time.

Sure.

Is the marking scale constructed by reading amc-compiled.pdf? Or one of the other files like amc-compiled.amc?

Not from the PDF.

maybe there's a way to have TeX expand the macros the document (writing to the .amc file in the process), but not write the PDF.

I think this would save a lot of time, but I did not find the way to do this (yet). If you find something, please tell!

Alexis Bienvenüe wrote:

maybe there's a way to have TeX expand the macros (in) the document (writing to the .amc file in the process), but not write the PDF.

I think this would save a lot of time, but I did not find the way to do this (yet). If you find something, please tell!

There is, but it doesn't. The syntonly package <http://www.ctan.org/pkg/syntonly>, which I just discovered this afternoon, does exactly that: process the document without writing any output. I tried putting this in the source file and regenerating the marking scale. It seemed to take the same amount of time for my large exam.

In the debugging file I noticed that the TeX process was still reading the graphics files cause by \includegraphics and \includepdf. So I added the "draft" option to both of those packages, and regenerated the marking scale. Then the included graphics files aren't read. Now the debugging file has almost nothing written to it except included TeX files and overfull hbox warnings. However, the job didn't finish any faster. It might even have been a little slower.

I am guessing that it's the pgfmath randomization that's the bottleneck. If so, perhaps using luatex instead of pgfmath would make it faster. I'll look into that next.

In the meantime, perhaps some user interface/documentation enhancements? The progress bar doesn't advance during TeXing. I guess it would be hard to make it do that since it's a subprocess. But it could easily be added to the documentation that regenerating the marking scale will take the same amount of time as the documents themselves. The same for the "compose" option on the Annotate function.