How do I get a single "closest" bookmark?

[lgarron]

As discussed in places like #2425 (reply in thread) , it can be useful to have a concept of a "closest" bookmark. However, all the syntax examples I can find (e.g. heads(::@- & bookmarks())) will resolve to multiple bookmarks sometimes.

Is there a way to get only the first of such a list? I've scoured the docs and tried a lot of syntax but I can't find a way to get it to give me a single one.

The closest I've found is to print them using bookmarks.map(|b| b.name() ++ \"\\n\") and pipe to head -n 1, but piping to head removes the color formatting. More importantly, it also prevents referring to a single bookmark in [aliases] in config.toml.

[scott2000]

You can use latest(heads(...)) to pick the most recent commit if there are multiple.

[lgarron]

Aha, looks like that works in the case I'm currently dealing with. Thanks!

[lgarron]

Hmm, I'm not so sure anymore. The documentation says:

latest(x[, count]): Latest count commits in x, based on committer timestamp. The default count is 1.

However, for my shell prompt I'd like to implement a way to get the closest bookmark with a fallback. In my case, I'd like to call back to fork_point(@ | trunk()).

What I "want" to do is the following, if first(…) existed:

first(coalesce(heads(::to & bookmarks()) | fork_point(@ | trunk())))

Replacing first(coalesce(…)) with latest(…) would prioritize the fork point over the bookmark in some cases. I guess what I'm looking for is:

• A way to give me the closest¹ change from a revset if it exists, else (but only if it does not exist) fall back to something else.
• Barring that, a way to get the "closest" in terms of commit distance, rather than latest (by commit date).

Is there a way to do either?

Footnotes

1. latest is fine in my case ↩

[scott2000]

I think coalesce(X, Y) is what you want here (not coalesce(X | Y), which is equivalent to X | Y). Maybe something like latest(coalesce(heads(::@ & bookmarks), fork_point(@ | trunk())))?

[lgarron]

Aha! That's super subtle, but I think it is indeed what I want. 🤩

[scott2000]

Yeah it's a bit confusing because for coalesce, you need separate arguments, but for fork_point, it just takes the single argument. I wonder if there should be an error or warning when using coalesce with only one argument.

[yuja]

In some circumstances, fork_point(x, y, ..) (where each argument may be multiple revisions) can be useful. We'll need a different function name or syntax, though.