allocate exact capacity for RFC 9110 strings

[zacknewman]

fmt::rfc2822::DateTimePrinter::timestamp_to_rfc9110_string allocates a String with capacity of 4; however RFC 9110-preferred-format strings are guaranteed to have an exact length of 29. I realize this crate doesn't prioritize performance; however the fact that String::with_capacity was already used as opposed to String::new suggests this isn't an entirely inappropriate diff.

[zacknewman]

The CI failure is obviously not related to this PR.

[BurntSushi]

Yeah this looks good!

FWIW, "not prioritizing performance" doesn't mean things shouldn't be optimized or fast or that Jiff doesn't care about being fast. The key word here is prioritization. Prioritization is about balance and what to do when multiple things are in tension with one another. For example, the Zoned and Span types both absolutely give up performance in some cases when compared to more spartan alternatives, but they provide more convenience and make it more difficult to shoot yourself in the foot.

For something like changing the initial capacity, there's really nothing that it is in tension with. And for sure, I have spent a fair bit of time optimizing aspects of Jiff. For example, I spent a lot of effort on an entire benchmark for one aspect of parsing the friendly duration format.

[zacknewman]

I didn’t mean anything by that comment other than its literal meaning. I chose “prioritize” and not “care” because as the maintainer of ripgrep, I know you care about performance. Perhaps I shouldn’t have said anything. Thanks for the quick merge.On Jan 12, 2025, at 11:50 AM, Andrew Gallant ***@***.***> wrote: @BurntSushi approved this pull request. Yeah this looks good! FWIW, "not prioritizing performance" doesn't mean things shouldn't be optimized or fast or that Jiff doesn't care about being fast. The key word here is prioritization. Prioritization is about balance and what to do when multiple things are in tension with one another. For example, the Zoned and Span types both absolutely give up performance in some cases when compared to more spartan alternatives, but they provide more convenience and make it more difficult to shoot yourself in the foot. For something like changing the initial capacity, there's really nothing that it is in tension with. And for sure, I have spent a fair bit of time optimizing aspects of Jiff. For example, I spent a lot of effort on an entire benchmark for one aspect of parsing the friendly duration format. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>