Every used template gets referenced exactly once

[Kijewski]

Once per #[template], that is. Still multiple times per crate.

Do you think that is enough to resolve #72, @GuillaumeGomez?

[Kijewski]

Feel free to tell me if you think this is overkill. I thought "lets make it pretty", but maybe I overdid it a bit :)

[GuillaumeGomez]

It's dev dependencies, so I definitely don't care. :)

[GuillaumeGomez]

I think it's definitely not worth it for "by template" approach. I also wondered about it myself but it really only makes sense for multiple templates. But as you underlined, the only solution for that would be tracked_file. I'll see what's missing for it to be done.

[Kijewski]

I think this diff might still make it worthwhile. Doesn't it?

[Kijewski]

Will be joined into one call with 3c08d0b.

[GuillaumeGomez]

Can you just provide a bench comparison in addition to the improved output please?

[Kijewski]

rinja_derive_standalone benchmark

hello_world             time:   [48.183 µs 48.253 µs 48.314 µs]
                        change: [-6.5054% -5.3650% -4.2704%] (p = 0.00 < 0.05)
                        Performance has improved.

item_info.html          time:   [58.699 µs 59.057 µs 59.467 µs]
                        change: [-1.5808% -0.9997% -0.4222%] (p = 0.00 < 0.05)
                        Change within noise threshold.

item_union.html         time:   [152.77 µs 152.92 µs 153.11 µs]
                        change: [-3.7432% -3.4421% -3.1460%] (p = 0.00 < 0.05)
                        Performance has improved.

page.html               time:   [651.01 µs 651.83 µs 652.81 µs]
                        change: [-3.8608% -3.5279% -3.1711%] (p = 0.00 < 0.05)
                        Performance has improved.

print_item.html         time:   [141.53 µs 142.07 µs 142.73 µs]
                        change: [-1.4684% -0.6647% +0.4653%] (p = 0.24 > 0.05)
                        No change in performance detected.

short_item_info.html    time:   [122.21 µs 122.49 µs 122.86 µs]
                        change: [-2.3561% -1.9451% -1.5023%] (p = 0.00 < 0.05)
                        Performance has improved.

sidebar.html            time:   [193.36 µs 195.57 µs 197.54 µs]
                        change: [+0.4147% +1.2141% +2.0348%] (p = 0.00 < 0.05)
                        Change within noise threshold.

source.html             time:   [105.18 µs 105.75 µs 106.48 µs]
                        change: [+1.3040% +2.5061% +3.6714%] (p = 0.00 < 0.05)
                        Performance has regressed.

type_layout.html        time:   [176.73 µs 176.98 µs 177.27 µs]
                        change: [-1.6070% -1.1707% -0.7469%] (p = 0.00 < 0.05)
                        Change within noise threshold.

type_layout_size.html   time:   [64.969 µs 65.371 µs 65.771 µs]
                        change: [-3.4571% -2.7019% -1.8661%] (p = 0.00 < 0.05)
                        Performance has improved.

IMO the results are spurious. We don't have a benchmark that uses included files, so the PR should not affect the results in one way or another, except that I replaced one call to quote! with format_args!.

[GuillaumeGomez]

What I was interested about was whether or not we had a perf regression. Since it's not the case, all good, thanks!