Documentation
¶
Overview ¶
Package history mines the git log into the history layer of the graph: commits become Decision rows, and file co-occurrence across commits becomes CoChange pairs with lift.
It shells out to git rather than using a Go git implementation: git is always present where a repo is, and `git log --numstat` is far faster than in-process history walking on large histories (RFC-001 §7.1).
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func FileCommits ¶
FileCommits returns the set of commit hashes that touched a file, capped to the mining window. It is the left side of a co-change intersection: the commits shared with a partner are those in BOTH files' sets. Empty (nil) when git is unavailable, so callers degrade to no enrichment.
func Mine ¶
Mine runs git log in repoRoot and returns decisions (all mined commits) plus co-change pairs (from commits passing the size filter). A repository with no commits yet yields empty results, not an error.
func PartnerFunctions ¶
PartnerFunctions reports which functions of partner were most often touched in the commits it shares with another file (shared). It streams the partner's diff history (git log -U0 -p, capped + timed out) and tallies the hunk-header function context of the shared commits only. This is a factual report of what moved together — never a statistical claim. NOTE: shared is the co-change window's commit set; a bulk commit dropped from the lift metric (too many files) that still touched both files can contribute here — a known, minor over-count.
Types ¶
type Options ¶
type Options struct {
// MaxCommits bounds the mined window. 0 means the default (5000).
// Co-change lift is computed over this window, so it reflects recent
// coupling rather than all-time archaeology.
MaxCommits int
// MaxFilesPerCommit drops bulk commits (formatting sweeps, vendoring,
// mass renames) from co-change counting; they say nothing about
// intentional coupling. 0 means the default (30). Dropped commits are
// still recorded as decisions.
MaxFilesPerCommit int
// MinTogether is the minimum number of shared commits for a pair to be
// stored. 0 means the default (2).
MinTogether int
}
Options tune the mining pass.
Source Files