latex woes

Stream: general

[Jeremy Tan]

If you had used Overleaf to write your paper you wouldn't have faced those build problems on the arXiv

[Jireh Loreaux]

I have a decent amount of experience bending arXiv to my will, if you want my help.

[Jeremy Tan]

I wrote my honours thesis on Zarankiewicz's problem over there, and it has an export-for-arXiv button

[Jeremy Tan]

https://arxiv.org/abs/2203.02283

[Eric Wieser]

I know exactly how to solve my problem but haven't gotten around to doing it because it's dull

[Kevin Buzzard]

whereas debugging typeclass issues for 90 minutes on Monday evening with me and Mario is fun!

[Anne Baanen]

Last time I opened Overleaf I got a message along the lines of "The source code editor is deprecated". I didn't realize the Embrace-Extend-Extinguish cycle had already landed on Extinguish...

[Anne Baanen]

(Do they have good Vim keybindings yet?)

[Eric Wieser]

The source code editor is deprecated

Well the previous one didn't support non-BMP unicode, which was rather damning for Lean anyway Nevermind, I mean the current one

[Jireh Loreaux]

Other issues with using Overleaf include:

1. I have to trust them with my data.
2. I'm locked in to whatever TeX they use / have available.
3. I have to be online in order to use it.

The first one is annoying, but the latter two are really deal breakers for me.

[Eric Wieser]

2 applies to arXiv and many journals too

[Moritz Doll]

in math you can ignore the instructions for journals usually (you have to deal with bad copy-editing anyways in the end) and for arXiv is really sad that they don't have support for something like luaTeX.

[Moritz Doll]

(there is a workaround by removing all the pdf-meta data and then uploading the pdf directly)

[Jireh Loreaux]

For 2 I'm referring to the development. If I have to coerce things at the end into a valid preprint or publishable format I'll do that. And for journals I've been successful in making them do what I want (e.g., adding new symbols that their LaTeX didn't have available or renumbering autonumbered things).

The worst outcome I ever had was that I wrote a (simple) macro which had a single optional argument with a default. Their processing did the strangest thing: if you used the macro without passing an argument it would display a ? where the argument was supposed to go ... until you called the macro and passed an optional argument, at which point that argument was "sticky". The net effect was that code which originally looked like this:

\macro
\macro[1]
\macro
\macro[2]
\macro

ended up producing an end result as if the source were this:

\macro[?]
\macro[1]
\macro[1]
\macro[2]
\macro[2]

This caused my innumerable headaches as the macro was used throughout the paper probably upwards of 200 times. Scouring their pdf and looking for all the inconsistencies was painstaking work.

[Jireh Loreaux]

(At this point we've digressed from the topic of this thread in case someone wants to move it)

[Notification Bot]

15 messages were moved here from #mathlib4 > splitting field discussion @ porting meeting by Heather Macbeth.

[Jannis Limperg]

I've started using Zenodo for preprints. None of this "you have to use LaTeX from 2001, also what is Unicode?" nonsense.

[Eric Wieser]

Moritz Doll said:

(there is a workaround by removing all the pdf-meta data and then uploading the pdf directly)

Do you have any tips for this? I've found various posts online about approaches that used to work but seem to no longer do so

[Eric Wieser]

I can use "Microsoft print to PDF which makes the file 10x bigger, which seems pretty stupid

[Eric Wieser]

But I guess if arXiv wants to force us to do that...