How to use arguments with hyphens in math mode

ian · December 7, 2024, 8:38pm

This is actually in a PR right now by mkorje!

github.com/typst/typst

Better math argument parsing

typst:main ← mkorje:better-math-argument-parsing

opened 05:07PM - 23 Sep 24 UTC

mkorje

+380 -121

This PR brings how math function call arguments are parsed more in line with how… they are parsed in code mode. New things: - Closes #2052: Named arguments in math function calls can now be any identifier. This means it can contain hyphens and start with an underscore (as long as it isn't just an underscore), just like in code mode. For example, the below now works. ```typ #let func(my-body: none) = my-body $ func(my-body: a) $ ``` - Closes #938: Argument spreading now works in math function calls. Prefixing two dots `..` to a variable does the same as in code mode. You can escape the dots or add a space after them to get `..` verbatim. See the example below. <table> <tr> <td> Code </td> <td> Image </td> </tr> <tr> <td> ```typ #set page(width: 5cm, height: auto) $ // Spreading the argument, so works. mat(..#range(1, 5).chunks(2)) // Passed verbatim, so reprs the list. mat(#(..range(2).map(_ => range(2)))) $ #let nums = ((1,) * 5).intersperse(0).chunks(3) $ mat(..nums, delim: "[") $ ``` </td> <td> ![math](https://github.com/user-attachments/assets/07362233-da6a-4aa8-a971-b21e892161b6) </td> </tr> </table> - Duplicate named arguments error: The below now gives an error, instead of silently overriding the previous instance of the named argument given. ```typ #let func(my: none) = my // Error: 15-17 duplicate argument: my $ func(my: a, my: b) $ ``` These changes don't conflict with #4364 majorly. Some things, like the new lex modes, might end up a bit cleaner with the changes from #4364 I think. Also, these changes are somewhat isolated; only a little (ish) has spread outside of math argument parsing to the rest of the math parsing code. Nothing else has changed with regards to math parsing (at least I hope so, all of the tests have passed which is promising). Only one test was updated to test `column-gap` and `row-gap` being used directly in matrices.