New module: X.L.ConditionalLayout

[slotThe]

Supersedes #582

What I haven't done, yet we probably should (not necessarily in this PR):

• Rewrite IfMax, PreScreen, and PerWorkspace with this, which would close Display bug with IfMax, PerScreen, PerWorkspace with decorated layouts  #75
• Provide some QOL functions so that it's easier to write predicates. E.g., I reckon lots of people will want to use a variant of gets $ find ((t ==) . W.tag) . W.workspaces . windowset.
• I'm not sure if this is enough docs still

TODOS:

• Find out what's up with New module: X.L.ConditionalLayout #825 (comment)
• Docs and changelog entries for all the deprecations
• Docs for the exported helper functions

New module: X.L.ConditionalLayout

Provide conditional variants of 'ModifiedLayout' and 'Choose', so that modifications (specific layouts) are only applied when a particular condition is met.

Co-authored-by: Tomas Janousek [email protected]
Co-authored-by: Ivan Malison [email protected]

---

Description

Include a description for your changes, including the motivation
behind them.

Checklist

• I've read CONTRIBUTING.md

• I've considered how to best test these changes (property, unit,
  manually, ...) and concluded: cursory testing on my computer yielded the desired results, though someone else should look at this.

• I updated the CHANGES.md file

[slotThe]

There's a but that I can't figure out right now: Assuming something like

myLayout = onWorkspace "4" (tall ||| Full) $ Full ||| tall ||| Mirror tall
 where
  tall = Tall 1 (3/100) (1/2)

one needs to issue five M-SPC calls to cycle through all layouts, regardless which branch one is on right now (some of those M-SPC invocations will just not do anything—at least not visibly).

Haven't tried debugging this yet, as its getting late, but perhaps someone else immediately knows what's up.

[geekosaur]

Is using Choose internally wise? That's going to add a couple extra mod-space-s and I'm wondering how that interacts with the explicit condition (wouldn't it effectively invert it?).

[geekosaur]

Also it may be better to leave PerScreen alone until we decide what's happening with #813.

[liskin]

I wonder if we really want this — this gives you two separate instances of l, so whenever the condition flips, you "lose" (as in, no longer observe, rather than losing the actual data) the internal state of l.

If l is something like l1 ||| l2 ||| l3, then the actual Chosen layout might switch every time the condition flips, and also the effect of Shrink/Expand/… only applies to the instance you're looking at.

So I'd say the implementation simplification this provides isn't worth it.

[liskin]

Also it may be better to leave PerScreen alone until we decide what's happening with #813.

I'm generally in favor of having everything implemented via CondChoose/CondModifiedLayout and making sure these two are perfect (as in, no resource leaks, no stuck decorations, etc.).

(Apologies for disappearing for a couple years. I've been extremely busy, and also extremely burned out. That's not getting better any time soon, but I finally reached a point of needing ConditionalLayout myself, so at least I have some motivation to progress this specific one…)