Expose Regex memory usage info (fixes #943)

[mispecto]

No description provided.

[ambyjkl]

@BurntSushi sorry for the ping but I wanted to bring your attention to this PR, it would be great if this could get merged

[ambyjkl]

@spektom can you also expose the function on RegexSet?

[BurntSushi]

I have concerns about this that I explained in #943. Those need to be resolved first.

[ambyjkl]

I have concerns about this that I explained in #943. Those need to be resolved first.

That's fair, maybe a better name would be memory_usage_estimate with some warnings about potential caveats. My use case doesn't need precise figures, I want to use this memory usage estimate as weights in a weighted cache, so all I need is the ratio of memory usage between two regexes, not an absolute value.

[LuisDuarte1]

I have almost the same usecase as @ambyjkl - I basically want a "LRU cache" of regex rules as I'm in a semi-constrained memory environment. In my opinion, calling it memory_usage_estimate is more than good enough providing we document it properly.

[typetetris]

Having a function like memory_usage_estimate for the exact same reason of having a size based cache would be very helpful to me, too.

At the moment I use regex-automata directly for that purpose and need to copy some stuff from the regex crate.

[ashdnazg]

I also have the exact same caching motivation.

@BurntSushi Any way to help push this beyond suggesting names? I think we'd all be happy with any of memory|heap|allocated with any of usage|bytes etc. (avoid using size to make it clear it is not about the size limit), add approximate|estimated before or estimate after, and that should be fine.

IMO best is approximate_allocated_bytes since even though it's quite the mouthful, it sounds like it has nothing to do with the size limit.

[BurntSushi]

I think memory_usage is probably good enough of a name. But the documentation should specifically call out that this isn't related to RegexBuilder::size_limit. That is, one cannot count on memory_usage() < RegexBuilder::size_limit().

[ashdnazg]

I opened a new PR #1377 since I can't update this one.