[DIT-13470] collect git file rename data

[jholiga]

Overview

CLI half of https://linear.app/dittowords/issue/DIT-13470/rename-pass. Before a scan starts, the CLI asks the server which commit the last scan read. It diffs that commit against HEAD with git diff -M and records what git says was moved or renamed.

Something to note is that git's similarity index decides what counts as a rename, not us. A heavily rewritten file is not a rename.

Every failure path returns no renames but never blocks the scan. That covers no git, no earlier scan, a shallow clone, an unreachable server, etc. A move of more than 1000 files is capped..

Context

• DIT-13470
• ditto-app PR: https://github.com/dittowords/ditto-app/pull/9443

Merge the ditto-app PR first.

Screenshots

Test Plan

• Ensure a ditto scan runs successfully. We can't test a re-scan until reconciliation is wired in.

[laurakoye]

Should we be logging here?

[jholiga]

going to skip the logging here, since this is a valid path for initial scans where no existing scan exists yet.

[laurakoye]

type the return

[laurakoye]

I believe this can be simplified. Adding --diff-filter=R makes git send back only the renames. Every record is then the same shape, so the parser becomes a step by 3 loop.

Worth it mainly for the output size.

Suggested change

	const out = await git(
	["diff", "-M", "--name-status", "-z", previousSha, "HEAD"],
	repoRoot
	);
	if (!out) return [];
	// `-z` gives NUL-separated fields: a status, then one path, or two for a rename or a copy.
	const fields = out.split("\0");
	const renames: GitRename[] = [];
	for (let i = 0; i < fields.length && fields[i]; ) {
	const status = fields[i][0];
	if (status === "R" || status === "C") {
	const [from, to] = [fields[i + 1], fields[i + 2]];
	if (status === "R" && from && to) renames.push({ from, to });
	i += 3;
	} else {
	i += 2;
	}
	}
	return renames;
	}
	const out = await git(
	["diff", "-M","--diff-filter=R","--name-status","-z",previousSha,"HEAD"],
	repoRoot
	);
	if (!out) return [];
	const fields = out.split("\0").filter(Boolean);
	const renames: GitRename[] = [];
	for (let i = 0; i + 2 < fields.length; i += 3) {
	renames.push({ from: fields[i + 1], to: fields[i + 2] });
	}
	return renames;

[jholiga]

nice catch!