Handling of short combing arrow ligatures such as >>= should be reconsidered

[schuelermine]

Environment

Cascadia Code version number: 2102.003
Application (with version) used to display text: konsole 20.08.2
Also occurs when using:
    Visual Studio Code 1.55.0-insider 84fe402d655e029eb1a5c04e675bf64788fa7fcf x64
    gedit - Version 3.38.0
OS platform and version: Pop!_OS 20.10
Screen resolution (i.e. 220dpi):
    1920px×1080px
    344mm×194mm (96x96dpi)

Steps to reproduce

Type >>=.

Expected behavior

A ligature appears, as with =<<.

Actual behavior

No ligature appears.

Screenshot

[aaronbell]

This is intentional.

=<< was implemented for the arbitrary-length arrows so you can write things like ==<<=== and so forth. However, in the case of >>= there is an operator in JavaScript that would be strange to replace with an arrow form. As a compromise, if you add another =, then it'll swap for the arrow ligature.

[schuelermine]

Sure, but Haskell uses this for the bind operator, and there, it's expected to be in ligature form, especially since =<< has the same meaning, but flipped.

[aaronbell]

Ah, well then perhaps I should disable =<< too then :)

[mfocko]

Ah, well then perhaps I should disable =<< too then :)

tbh if there are more suggestions for Haskell ligatures, you could turn it into stylistic set, right?

[aaronbell]

Certainly could.

[aaronbell]

I think for now I'll disable arrow forms for all of the associated operators (=>> <<= =<< and >>=), but might add special ligatures for them in the future to give a bit of a nicer appearance—or as @mfocko recommended, add a stylistic set for Haskell use. Thanks!

[DHowett]

Aw, I love the arrow forms.

[aaronbell]

you can still get to them via ==>> <<== ==<< and >>==! Honestly, I think it would be nice to have ligatures for these, I just don't like the inconsistency :/

[schuelermine]

you can still get to them via ==>> <<== ==<< and >>==! Honestly, I think it would be nice to have ligatures for these, I just don't like the inconsistency :/

I think that line of reasoning is misguided. The thing we like about the arrow forms is that they appear in our code, not that we can type them.

[aaronbell]

I don't disagree that it is nice to see them in Haskell. The problem is that they cause confusion / irritation in other scenarios and users of those scenarios might prefer to not have them look like arrows as they serve other purposes.

This is why I said that it would either make sense to create non-arrow ligature forms or provide a stylistic set for Haskell-only functionality. Either way, these operators should perform the same.

[schuelermine]

I agree. I just disagree with the idea that the fact that similar ligatures are still available diminishes the Haskell use case.

I really like the idea of a stylistic set for Haskell, I think that would be the best option. However, why not both? A JS-style ligature, and a Haskell-style ligature.

[aaronbell]

Oh, no I didn't mean to diminish the Haskell use case. More to indicate that the arrows are still available if one wants to make arrows. :)

Anyway, I'll mull over options.

[schuelermine]

Cool, I'm looking forward to any solutions!
I changed the title to reflect that the topic has widened.

[ExE-Boss]

I think it might be a good compromise to create a ligature for all of: >>=, =>>, <<= and =<<, that uses a narrower version of the >> and << ligatures as a base, with the = sign extending from it.

---

There currently isn’t any ligature for either of >>= and <<=, which is used as a bit‑shift and assign operator in C‑like languages (maybe also include the unsigned variant: >>>= and <<<=).

[chtenb]

These ligatures don't only apply for Haskell, but also for PureScript. In fact many functional languages using operators heavily are relevant for this.

[2jacobtan]

rip haskell users, since https://github.com/microsoft/cascadia-code/releases/tag/v2106.17

really liked the ligatures on >>= and =<<

(No discernible ligature, using the latest release https://github.com/microsoft/cascadia-code/releases/tag/v2111.01)