ruff/scripts/update_ambiguous_characters.py
konstin 651d89794c
Use phf for confusables to reduce llvm lines (#4926)
* Use phf for confusables to reduce llvm lines

## Summary

This replaces FxHashMap for the confusables with a perfect hash map from the [phf crate](https://github.com/rust-phf/rust-phf) to reduce the generated llvm instructions.

A perfect hash function is one that doesn't have any collisions. We can build one because we know all keys at compile time. This improves hashmap efficiency, even though this is likely not noticeable in our case (except someone has a large non-english crate to test on).

The original hashmap contained a lot of duplicates, which i had to remove when phf_map complained, i did so by sorting the keys.

The important part that it reduces the llvm instructions generated (#3808, `RUSTFLAGS="-Csymbol-mangling-version=v0" cargo llvm-lines -p ruff --lib | head -20`):

```
  Lines                  Copies               Function name
  -----                  ------               -------------
  1740502                38973                (TOTAL)
    27423 (1.6%,  1.6%)      1 (0.0%,  0.0%)  ruff[cef4c65d96248843]::rules::ruff::rules::confusables::CONFUSABLES::{closure#0}
    10193 (0.6%,  2.2%)      1 (0.0%,  0.0%)  <ruff[cef4c65d96248843]::codes::RuleCodePrefix>::iter
     8107 (0.5%,  2.6%)      1 (0.0%,  0.0%)  <ruff[cef4c65d96248843]::codes::Rule>::noqa_code
     7345 (0.4%,  3.0%)      1 (0.0%,  0.0%)  <ruff[cef4c65d96248843]::checkers::ast::Checker as ruff_python_ast[3778b140caf21545]::visitor::Visitor>::visit_stmt
     6412 (0.4%,  3.4%)      1 (0.0%,  0.0%)  <<ruff[cef4c65d96248843]::settings::options::Options as serde[d89b1b632568f5a3]:🇩🇪:Deserialize>::deserialize::__Visitor as serde[d89b1b632568f5a3]:🇩🇪:Visitor>::visit_map::<toml_edit[7e3a6c5e67260672]:🇩🇪:spanned::SpannedDeserializer<toml_edit[7e3a6c5e67260672]:🇩🇪:value::ValueDeserializer>>
     6412 (0.4%,  3.8%)      1 (0.0%,  0.0%)  <<ruff[cef4c65d96248843]::settings::options::Options as serde[d89b1b632568f5a3]:🇩🇪:Deserialize>::deserialize::__Visitor as serde[d89b1b632568f5a3]:🇩🇪:Visitor>::visit_map::<toml_edit[7e3a6c5e67260672]:🇩🇪:table::TableMapAccess>
     6409 (0.4%,  4.2%)      1 (0.0%,  0.0%)  <<ruff[cef4c65d96248843]::settings::options::Options as serde[d89b1b632568f5a3]:🇩🇪:Deserialize>::deserialize::__Visitor as serde[d89b1b632568f5a3]:🇩🇪:Visitor>::visit_map::<toml_edit[7e3a6c5e67260672]:🇩🇪:datetime::DatetimeDeserializer>
     5696 (0.3%,  4.5%)      1 (0.0%,  0.0%)  <ruff[cef4c65d96248843]::checkers::ast::Checker as ruff_python_ast[3778b140caf21545]::visitor::Visitor>::visit_expr
     4448 (0.3%,  4.7%)      1 (0.0%,  0.0%)  ruff[cef4c65d96248843]::flake8_to_ruff::converter::convert
     3702 (0.2%,  4.9%)      1 (0.0%,  0.0%)  <&ruff[cef4c65d96248843]::registry::Linter as core[da82827a87f140f9]::iter::traits::collect::IntoIterator>::into_iter
     3349 (0.2%,  5.1%)      1 (0.0%,  0.0%)  <ruff[cef4c65d96248843]::registry::Linter>::code_for_rule
     3132 (0.2%,  5.3%)      1 (0.0%,  0.0%)  <ruff[cef4c65d96248843]::codes::Rule as core[da82827a87f140f9]::fmt::Debug>::fmt
     3130 (0.2%,  5.5%)      1 (0.0%,  0.0%)  <&str as core[da82827a87f140f9]::convert::From<&ruff[cef4c65d96248843]::codes::Rule>>::from
     3130 (0.2%,  5.7%)      1 (0.0%,  0.0%)  <&str as core[da82827a87f140f9]::convert::From<ruff[cef4c65d96248843]::codes::Rule>>::from
     3130 (0.2%,  5.9%)      1 (0.0%,  0.0%)  <ruff[cef4c65d96248843]::codes::Rule as core[da82827a87f140f9]::convert::AsRef<str>>::as_ref
     3128 (0.2%,  6.0%)      1 (0.0%,  0.0%)  <ruff[cef4c65d96248843]::codes::RuleIter>::get
     2669 (0.2%,  6.2%)      1 (0.0%,  0.0%)  <<ruff[cef4c65d96248843]::settings::options::Options as serde[d89b1b632568f5a3]:🇩🇪:Deserialize>::deserialize::__Visitor as serde[d89b1b632568f5a3]:🇩🇪:Visitor>::visit_seq::<toml_edit[7e3a6c5e67260672]:🇩🇪:array::ArraySeqAccess>
```
After:
```
  Lines                  Copies               Function name
  -----                  ------               -------------
  1710487                38900                (TOTAL)
    10193 (0.6%,  0.6%)      1 (0.0%,  0.0%)  <ruff[52408f46d2058296]::codes::RuleCodePrefix>::iter
     8107 (0.5%,  1.1%)      1 (0.0%,  0.0%)  <ruff[52408f46d2058296]::codes::Rule>::noqa_code
     7345 (0.4%,  1.5%)      1 (0.0%,  0.0%)  <ruff[52408f46d2058296]::checkers::ast::Checker as ruff_python_ast[5588cd60041c8605]::visitor::Visitor>::visit_stmt
     6412 (0.4%,  1.9%)      1 (0.0%,  0.0%)  <<ruff[52408f46d2058296]::settings::options::Options as serde[d89b1b632568f5a3]:🇩🇪:Deserialize>::deserialize::__Visitor as serde[d89b1b632568f5a3]:🇩🇪:Visitor>::visit_map::<toml_edit[7e3a6c5e67260672]:🇩🇪:spanned::SpannedDeserializer<toml_edit[7e3a6c5e67260672]:🇩🇪:value::ValueDeserializer>>
     6412 (0.4%,  2.2%)      1 (0.0%,  0.0%)  <<ruff[52408f46d2058296]::settings::options::Options as serde[d89b1b632568f5a3]:🇩🇪:Deserialize>::deserialize::__Visitor as serde[d89b1b632568f5a3]:🇩🇪:Visitor>::visit_map::<toml_edit[7e3a6c5e67260672]:🇩🇪:table::TableMapAccess>
     6409 (0.4%,  2.6%)      1 (0.0%,  0.0%)  <<ruff[52408f46d2058296]::settings::options::Options as serde[d89b1b632568f5a3]:🇩🇪:Deserialize>::deserialize::__Visitor as serde[d89b1b632568f5a3]:🇩🇪:Visitor>::visit_map::<toml_edit[7e3a6c5e67260672]:🇩🇪:datetime::DatetimeDeserializer>
     5696 (0.3%,  3.0%)      1 (0.0%,  0.0%)  <ruff[52408f46d2058296]::checkers::ast::Checker as ruff_python_ast[5588cd60041c8605]::visitor::Visitor>::visit_expr
     4448 (0.3%,  3.2%)      1 (0.0%,  0.0%)  ruff[52408f46d2058296]::flake8_to_ruff::converter::convert
     3702 (0.2%,  3.4%)      1 (0.0%,  0.0%)  <&ruff[52408f46d2058296]::registry::Linter as core[da82827a87f140f9]::iter::traits::collect::IntoIterator>::into_iter
     3349 (0.2%,  3.6%)      1 (0.0%,  0.0%)  <ruff[52408f46d2058296]::registry::Linter>::code_for_rule
     3132 (0.2%,  3.8%)      1 (0.0%,  0.0%)  <ruff[52408f46d2058296]::codes::Rule as core[da82827a87f140f9]::fmt::Debug>::fmt
     3130 (0.2%,  4.0%)      1 (0.0%,  0.0%)  <&str as core[da82827a87f140f9]::convert::From<&ruff[52408f46d2058296]::codes::Rule>>::from
     3130 (0.2%,  4.2%)      1 (0.0%,  0.0%)  <&str as core[da82827a87f140f9]::convert::From<ruff[52408f46d2058296]::codes::Rule>>::from
     3130 (0.2%,  4.4%)      1 (0.0%,  0.0%)  <ruff[52408f46d2058296]::codes::Rule as core[da82827a87f140f9]::convert::AsRef<str>>::as_ref
     3128 (0.2%,  4.5%)      1 (0.0%,  0.0%)  <ruff[52408f46d2058296]::codes::RuleIter>::get
     2669 (0.2%,  4.7%)      1 (0.0%,  0.0%)  <<ruff[52408f46d2058296]::settings::options::Options as serde[d89b1b632568f5a3]:🇩🇪:Deserialize>::deserialize::__Visitor as serde[d89b1b632568f5a3]:🇩🇪:Visitor>::visit_seq::<toml_edit[7e3a6c5e67260672]:🇩🇪:array::ArraySeqAccess>
     2659 (0.2%,  4.9%)      1 (0.0%,  0.0%)  <&ruff[52408f46d2058296]::codes::Pylint as core[da82827a87f140f9]::iter::traits::collect::IntoIterator>::into_iter
```

I'd assume this has a positive effect both on compile time and on runtime, but i don't know the actual effect on compile times and can't really measure.

## Test plan

Check CI for any performance regressions.

This should fix #3808 if we merge it.

* clippy

* Update update_ambiguous_characters.py
2023-06-08 08:13:20 +02:00

67 lines
2.3 KiB
Python

"""Generate the confusables.rs file from the VS Code ambiguous.json file."""
import json
import subprocess
from pathlib import Path
CONFUSABLES_RS_PATH = "crates/ruff/src/rules/ruff/rules/confusables.rs"
AMBIGUOUS_JSON_URL = "https://raw.githubusercontent.com/hediet/vscode-unicode-data/main/out/ambiguous.json"
prelude = """
/// This file is auto-generated by `scripts/update_ambiguous_characters.py`.
use phf::phf_map;
/// Via: <https://github.com/hediet/vscode-unicode-data/blob/main/out/ambiguous.json>
/// See: <https://github.com/microsoft/vscode/blob/095ddabc52b82498ee7f718a34f9dd11d59099a8/src/vs/base/common/strings.ts#L1094>
#[allow(clippy::unreadable_literal)]
pub(crate) static CONFUSABLES: phf::Map<u32, u8> = phf_map! {
""".lstrip()
postlude = """};"""
def get_mapping_data() -> dict:
"""
Get the ambiguous character mapping data from the vscode-unicode-data repository.
Uses the system's `curl` command to download the data,
instead of adding a dependency to a Python-native HTTP client.
"""
content = subprocess.check_output(
["curl", "-sSL", AMBIGUOUS_JSON_URL],
encoding="utf-8",
)
# The content is a JSON object literal wrapped in a JSON string, so double decode:
return json.loads(json.loads(content))
def format_confusables_rs(raw_data: dict[str, list[int]]) -> str:
"""Format the downloaded data into a Rust source file."""
# The input data contains duplicate entries
flattened_items: set[tuple[int, int]] = set()
for _category, items in raw_data.items():
assert len(items) % 2 == 0, "Expected pairs of items"
for i in range(0, len(items), 2):
flattened_items.add((items[i], items[i + 1]))
tuples = []
for left, right in sorted(flattened_items):
tuples.append(f" {left}u32 => {right},\n")
print(f"{len(tuples)} confusable tuples.")
return prelude + "".join(tuples) + postlude
def main() -> None:
print("Retrieving data...")
mapping_data = get_mapping_data()
formatted_data = format_confusables_rs(mapping_data)
confusables_path = Path(__file__).parent.parent / CONFUSABLES_RS_PATH
confusables_path.write_text(formatted_data, encoding="utf-8")
print("Formatting Rust file with cargo fmt...")
subprocess.check_call(["cargo", "fmt", "--", confusables_path])
print("Done.")
if __name__ == "__main__":
main()