evo grep never uses ripgrep — availability check is inverted, always falls back to grep #94

Summary

evo grep is documented/intended to prefer ripgrep and fall back to grep, but the availability check is inverted, so rg is never selected — every evo grep runs grep -rn.

Location

plugins/evo/src/evo/cli.py:2432 (cmd_ws_grep):

which = executor.run(["which", "rg"], cwd=str(base))
if (which.exit_code or 1) == 0 and which.stdout.strip():
    cmd = ["rg", "--no-heading", "--line-number", args.pattern, str(base)]
else:
    cmd = ["grep", "-rn", args.pattern, str(base)]

Root cause

which rg exits 0 when rg is present. But (which.exit_code or 1) collapses the success value: 0 is falsy, so 0 or 1 → 1, and the test becomes 1 == 0 → False. When rg is absent the exit code is 1, so 1 or 1 → 1, again 1 == 0 → False.

The condition is False in every case → the rg branch is dead code. The or 1 idiom (meant to coerce a None exit code into a failure sentinel) also clobbers the only value that should pass the check.

>>> (0 or 1) == 0   # rg present
False
>>> (1 or 1) == 0   # rg absent
False

Impact

On any machine with rg installed, evo grep <pattern> silently runs grep -rn instead. Because rg and grep differ in:

• regex dialect — a pattern written for rg (\d, \b, lookarounds) matches differently or errors under BRE grep;
• ignore semantics — rg respects .gitignore (skips node_modules/, build dirs); grep -rn searches them.

…the tool returns different/wrong results than intended, and the "prefer rg" contract is entirely non-functional.

Proposed fix

if which.exit_code == 0 and which.stdout.strip():

(or, preserving the None-guard intent: if (which.exit_code if which.exit_code is not None else 1) == 0 and which.stdout.strip():).

Verify

Stub executor.run(["which","rg"]) → exit_code=0, stdout="/usr/bin/rg\n" and assert cmd[0] == "rg"; it fails on current code. Or add a trace print of cmd and run evo grep foo in a workspace with rg on PATH — it picks grep.