관리-도구
편집 파일: raw_metrics.py
# Licensed under the GPL: https://www.gnu.org/licenses/old-licenses/gpl-2.0.html # For details: https://github.com/PyCQA/pylint/blob/main/LICENSE # Copyright (c) https://github.com/PyCQA/pylint/blob/main/CONTRIBUTORS.txt from __future__ import annotations import sys import tokenize from typing import TYPE_CHECKING, Any, cast from pylint.checkers import BaseTokenChecker from pylint.reporters.ureports.nodes import Paragraph, Section, Table, Text from pylint.utils import LinterStats, diff_string if sys.version_info >= (3, 8): from typing import Literal else: from typing_extensions import Literal if TYPE_CHECKING: from pylint.lint import PyLinter def report_raw_stats( sect: Section, stats: LinterStats, old_stats: LinterStats | None, ) -> None: """Calculate percentage of code / doc / comment / empty.""" total_lines = stats.code_type_count["total"] sect.insert(0, Paragraph([Text(f"{total_lines} lines have been analyzed\n")])) lines = ["type", "number", "%", "previous", "difference"] for node_type in ("code", "docstring", "comment", "empty"): node_type = cast(Literal["code", "docstring", "comment", "empty"], node_type) total = stats.code_type_count[node_type] percent = float(total * 100) / total_lines if total_lines else None old = old_stats.code_type_count[node_type] if old_stats else None diff_str = diff_string(old, total) if old else None lines += [ node_type, str(total), f"{percent:.2f}" if percent is not None else "NC", str(old) if old else "NC", diff_str if diff_str else "NC", ] sect.append(Table(children=lines, cols=5, rheaders=1)) class RawMetricsChecker(BaseTokenChecker): """Checker that provides raw metrics instead of checking anything. Provides: * total number of lines * total number of code lines * total number of docstring lines * total number of comments lines * total number of empty lines """ # configuration section name name = "metrics" # configuration options options = () # messages msgs: Any = {} # reports reports = (("RP0701", "Raw metrics", report_raw_stats),) def open(self) -> None: """Init statistics.""" self.linter.stats.reset_code_count() def process_tokens(self, tokens: list[tokenize.TokenInfo]) -> None: """Update stats.""" i = 0 tokens = list(tokens) while i < len(tokens): i, lines_number, line_type = get_type(tokens, i) self.linter.stats.code_type_count["total"] += lines_number self.linter.stats.code_type_count[line_type] += lines_number JUNK = (tokenize.NL, tokenize.INDENT, tokenize.NEWLINE, tokenize.ENDMARKER) def get_type( tokens: list[tokenize.TokenInfo], start_index: int ) -> tuple[int, int, Literal["code", "docstring", "comment", "empty"]]: """Return the line type : docstring, comment, code, empty.""" i = start_index start = tokens[i][2] pos = start line_type = None while i < len(tokens) and tokens[i][2][0] == start[0]: tok_type = tokens[i][0] pos = tokens[i][3] if line_type is None: if tok_type == tokenize.STRING: line_type = "docstring" elif tok_type == tokenize.COMMENT: line_type = "comment" elif tok_type in JUNK: pass else: line_type = "code" i += 1 if line_type is None: line_type = "empty" elif i < len(tokens) and tokens[i][0] == tokenize.NEWLINE: i += 1 # Mypy fails to infer the literal of line_type return i, pos[0] - start[0] + 1, line_type # type: ignore[return-value] def register(linter: PyLinter) -> None: linter.register_checker(RawMetricsChecker(linter))