gh-118750: Asymptotically faster int(string)

[tim-one]

Adding new, but unsued, _dec_str_to_int_inner(), + discussion.

• Issue: Asymptotically faster int(string) #118750

[nineteendo]

Skip news?

[tim-one]

Skip news?

Premature, I think. It's possible that a later version may become actively used.

[nineteendo]

We only need a news entry for user facing changes. When the faster integer conversion is eventually implemented, you'll be credited.

[tim-one]

Yes, I understand that. I'm saying that it's possible this will become a "user-facing" change soon. I already have a newer version (not yet checked in) that's much better. Mucking with labels is, IMO, a minor waste of time at this point.

[nineteendo]

Sorry, I'm just trying to help. I thought you didn't mean in the near future.

[tim-one]

No problem! Things change: I originally thought this was far from being ready, but that's been changing rapidly.

[tim-one]

A cute mystery: I noticed that converting a string representing an int a little smaller than a power of 256 took about 3x longer than one a little larger than that power. Turns out there's "a reason" for that: in the latter case, the n inner() sees is frequently 0, so it cuts back to a mere 8 digits of precision to compute the integer quotient. In the former case, n is usually chock full o' 1 bits, so the precision is only cut in half.

[tim-one]

Dear Lord, buildbots can be annoying 😉.

It's hard to be sure of things digging through the mountains of text files they save. Best guess is the failing boxes do not build the C _decimal module, so mostly (always?) time out running test_int. I took my best guess at that and changed the tests most likely to time out then to not run at all unless _decimal is available, and checked that in.

But most of the buildbots that failed are just sitting there, apparently doing nothing, and trying to "force" a new run just says the request is queued - but not running it.

Just 3 of these buildbots appeared to wake up and actually run again after that commit. Their tests all passed.

So my best guess is that the problem is fixed. But no way yet to be sure.