class

Analyzer::Python::Cli

Inherits CliEndpointSupport < Analyzer::Python::PythonEngine < Analyzer < FileHelper < Reference < Object

Surfaces the command-line attack surface of Python programs as cli:// endpoints: one endpoint per (sub)command, with named options (param_type "flag"), positional arguments ("argument"), and consumed environment variables ("env"). Covers stdlib argparse / getopt / sys.argv plus click, typer, fire, docopt, Abseil (absl-py) flags, and Cleo commands.

Line-scan analyzer (the house style for non-tree-sitter Python adapters, e.g. bottle). Endpoints are merged by URL so options registered across decorators/functions collect onto a single command.

Constants

ABSL_DEFINE_CALL_RE = /\bflags\.DEFINE_(?:string|integer|bool|boolean|float|enum|list|multi_string|multi_integer|multi_enum)\s*\(/

Abseil (absl-py): flags.DEFINE_* calls register a single flat flag namespace consumed via FLAGS.<name> after app.run(main) — there is no subcommand concept, so every flag attaches to the binary's root URL.

ABSL_DEFINE_NAME_RE = /\bflags\.DEFINE_(?:string|integer|bool|boolean|float|enum|list|multi_string|multi_integer|multi_enum)\s*\(\s*[rf]?["']([^"']+)["']/
ADD_ARGUMENT_RE = /(\w+)\s*\.\s*add_argument\s*\(([^)]*)/
ADD_PARSER_RE = /(\w+)\s*=\s*(\w+)\.add_parser\s*\(\s*[rf]?["']([^"']+)["']/
ARGPARSE_NEW_RE = /(\w+)\s*=\s*argparse\.ArgumentParser\s*\(/

argparse

ARGPARSE_PROG_RE = /ArgumentParser\([^)]*\bprog\s*=\s*[rf]?["']([^"']+)["']/
CLEO_ARGUMENT_CALL_RE = /\bself\.argument\s*\(\s*[rf]?["']([^"']+)["']/
CLEO_COMMAND_CLASS_ANYWHERE_RE = /(?:^|\n)class\s+\w+\s*\(\s*Command\b/
CLEO_COMMAND_CLASS_RE = /^class\s+(\w+)\s*\(\s*Command\b/

^ in Crystal/PCRE without the multiline flag anchors to the start of the whole subject, not each line — fine for the per-line line.match use in scan_cleo below, but a separate (?:^|\n)-anchored variant is needed to test this against the full multi-line file content. Both variants are column-0 anchored (no \s* before class) so the entrypoint gate below matches exactly what scan_cleo can extract — a class Foo(Command) nested inside a function/method (indented) is never resolved by scan_cleo, so it must not satisfy the gate either, or the file gets treated as a CLI entrypoint with no real command found (scan_stdlib then fires unconditionally on an unrelated env/argv read).

CLEO_IMPORT_RE = /(?:^|\n)\s*(?:import\s+cleo\b|from\s+cleo(?:\.\w+)*\s+import\b)/

Cleo: a Command subclass declares its subcommand name as a class attribute, then reads its own arguments/options back out by name inside its methods (self.argument("x") / self.option("y")) — that readback call is also the most reliable textual anchor for the option/argument name itself.

CLEO_NAME_ATTR_RE = /^\s*name\s*=\s*[rf]?["']([^"']+)["']/
CLEO_OPTION_CALL_RE = /\bself\.option\s*\(\s*[rf]?["']([^"']+)["']/
CLI_LITERAL_MARKERS_RE = Regex.union("argparse.ArgumentParser", "click.command", "click.group", "typer.Typer", "fire.Fire", "getopt.getopt", "docopt")

The seven literal markers below were seven separate Rabin-Karp passes over the whole file, run for every .py in the tree. One precompiled union is the same predicate in a single pass. The remaining tests go through content_matches? so PCRE2 skips its per-call UTF-8 revalidation of the buffer. The gate is a pure OR, so collapsing and reordering its terms cannot change the answer.

CLICK_ARGUMENT_RE = /^@\w+\.argument\s*\(/
CLICK_OPTION_RE = /^@\w+\.option\s*\(/
DECORATOR_CMD_RE = /^@(\w+)\.(command|group|callback)\s*\(/

click / typer decorators

DECORATOR_NAME_RE = /\(\s*[rf]?["']([^"']+)["']/
DEF_RE = /^def\s+(\w+)\s*\(/
ENVVAR_RE = /\benvvar\s*=\s*[rf]?["']([^"']+)["']/
FIRE_RE = /\bfire\.Fire\s*\(\s*(\w+)/
GETOPT_RE = /getopt\.getopt\s*\([^,]+,\s*[rf]?["']([^"']*)["']\s*(?:,\s*\[([^\]]*)\])?/
MAIN_GUARD_RE = /if\s+__name__\s*==\s*[rf]?["']__main__["']/
OS_ENVIRON_GET_RE = /\bos\.environ\.get\s*\(\s*[rf]?["']([^"']+)["']/
OS_ENVIRON_RE = /\bos\.environ\s*\[\s*[rf]?["']([^"']+)["']\s*\]/
OS_GETENV_RE = /\bos\.getenv\s*\(\s*[rf]?["']([^"']+)["']/
SYS_ARGV_RE = /\bsys\.argv\s*\[\s*(\d+)\s*\]/

stdlib argv / env / getopt

TYPER_ARGUMENT_PARAM_RE = /(\w+)\s*:[^,=]+=\s*typer\.Argument/
TYPER_NEW_RE = /(\w+)\s*=\s*typer\.Typer\s*\(/

typer

TYPER_OPTION_PARAM_RE = /(\w+)\s*:[^,=]+=\s*typer\.Option/
WEB_FRAMEWORK_RE = /(?:^|\n)\s*(?:import|from)\s+(?:flask|django|fastapi|starlette|sanic|aiohttp|bottle|tornado|falcon|pyramid|quart|litestar|robyn)\b/

Web frameworks: their os.environ/os.getenv reads are config, not a CLI surface, so raw env is suppressed when one is present (framework-bound env via click/typer is still emitted).

Class methods

tech_name
Source

Instance methods

analyze
Source
tech

Instance-side view of the same declaration. The per-file rescues live on this base class, which has no way to name the analyzer that is running inside them, so a skipped file could not be attributed to a tech. Deriving it from analyzer_for keeps the name written exactly once.

Source