Printer: Reserve buffer upfront

[MichaReiser]

Summary

I did some profiling on the formatter and noticed that we end up growing the Printer buffer multiple times. This PR tries to reduce the number of times that we need to grow the formatted string buffer by allocating a buffer of the same size as the unformatted document.
This should be a good default because most files in a project are already formatted (we neither allocate too much nor too little).

Test Plan

This improved the formatter time by about 2%

cargo test

[MichaReiser]

Current dependencies on/for this PR:

• main
  • PR Remove lex and parsing from formatter benchmark #6547
    • PR Printer: Remove unused state fields #6548
      • PR Remove allow(pedantic) from formatter #6549
        • PR Printer: Reserve buffer upfront #6550 👈
          • PR Memoize text width #6552
          • PR Override fmt_dangling_comments for frequent nodes #6551

This comment was auto-generated by Graphite.

[MichaReiser]

It may be interesting to see if we can come up with a good heuristic for the VecBuffer size. E.g. is it always around ~0.7 * the source code (@konstin ;))

[github-actions]

PR Check Results

Benchmark

Linux

group                                      main                                   pr
-----                                      ----                                   --
formatter/large/dataset.py                 1.00      3.9±0.02ms    10.3 MB/sec    1.00      3.9±0.03ms    10.4 MB/sec
formatter/numpy/ctypeslib.py               1.00    719.0±8.14µs    23.2 MB/sec    1.00    720.6±5.97µs    23.1 MB/sec
formatter/numpy/globals.py                 1.00     71.5±0.29µs    41.3 MB/sec    1.00     71.3±0.31µs    41.4 MB/sec
formatter/pydantic/types.py                1.00  1569.0±39.62µs    16.3 MB/sec    1.00  1573.4±37.73µs    16.2 MB/sec
linter/all-rules/large/dataset.py          1.00     10.6±0.04ms     3.8 MB/sec    1.01     10.7±0.04ms     3.8 MB/sec
linter/all-rules/numpy/ctypeslib.py        1.00      2.9±0.01ms     5.7 MB/sec    1.00      2.9±0.01ms     5.7 MB/sec
linter/all-rules/numpy/globals.py          1.00    330.4±2.27µs     8.9 MB/sec    1.00    329.5±1.44µs     9.0 MB/sec
linter/all-rules/pydantic/types.py         1.00      5.5±0.02ms     4.7 MB/sec    1.00      5.5±0.04ms     4.6 MB/sec
linter/default-rules/large/dataset.py      1.00      5.6±0.02ms     7.2 MB/sec    1.00      5.6±0.03ms     7.2 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.00   1190.0±2.92µs    14.0 MB/sec    1.00   1187.8±3.12µs    14.0 MB/sec
linter/default-rules/numpy/globals.py      1.00    124.7±0.30µs    23.7 MB/sec    1.00    124.6±0.34µs    23.7 MB/sec
linter/default-rules/pydantic/types.py     1.00      2.5±0.01ms    10.2 MB/sec    1.00      2.5±0.01ms    10.1 MB/sec

Windows

group                                      main                                   pr
-----                                      ----                                   --
formatter/large/dataset.py                 1.00      4.3±0.15ms     9.4 MB/sec    1.00      4.3±0.02ms     9.5 MB/sec
formatter/numpy/ctypeslib.py               1.00    781.8±7.25µs    21.3 MB/sec    1.01    787.8±8.73µs    21.1 MB/sec
formatter/numpy/globals.py                 1.00     78.6±0.67µs    37.5 MB/sec    1.01     79.4±1.22µs    37.2 MB/sec
formatter/pydantic/types.py                1.00  1759.5±19.87µs    14.5 MB/sec    1.00  1767.5±24.28µs    14.4 MB/sec
linter/all-rules/large/dataset.py          1.00     12.7±0.19ms     3.2 MB/sec    1.00     12.7±0.12ms     3.2 MB/sec
linter/all-rules/numpy/ctypeslib.py        1.00      3.5±0.01ms     4.8 MB/sec    1.00      3.5±0.02ms     4.8 MB/sec
linter/all-rules/numpy/globals.py          1.01    362.4±5.35µs     8.1 MB/sec    1.00    357.8±4.79µs     8.2 MB/sec
linter/all-rules/pydantic/types.py         1.00      6.5±0.05ms     3.9 MB/sec    1.00      6.5±0.05ms     3.9 MB/sec
linter/default-rules/large/dataset.py      1.00      6.9±0.03ms     5.9 MB/sec    1.01      6.9±0.05ms     5.9 MB/sec
linter/default-rules/numpy/ctypeslib.py    1.00   1427.9±9.89µs    11.7 MB/sec    1.00   1431.2±9.36µs    11.6 MB/sec
linter/default-rules/numpy/globals.py      1.00    142.7±1.16µs    20.7 MB/sec    1.00    142.7±1.93µs    20.7 MB/sec
linter/default-rules/pydantic/types.py     1.00      3.1±0.02ms     8.3 MB/sec    1.00      3.1±0.02ms     8.3 MB/sec

[konstin]

nice!

what i really want to benchmark is how many % extra we need to cover >95% reformattings too