Skip to main content

Automatically exclude cherry-pick, revert and rebase from statistics

An enabled Off/On toggle makes commits recognized as a cherry-pick, revert or rebase not count toward Calories and Lines, so the same work (moved from another branch, reverted, recreated) is not counted toward statistics more than once.

How Q247 recognizes these commits​

Every commit gets one of four classifications: normal, revert, cherry-pick or rebase duplicate. Recognition works independently of the toggle, so the classification is always recorded; the toggle only decides whether a duplicate drops out of statistics.

ClassificationWhat the recognition is based on
Revertthe line This reverts commit <identifier>, which Git appends to the message on git revert. It is the message body that counts, not its first line.
Cherry-pickthe annotation cherry picked from commit <identifier>, which Git appends on git cherry-pick -x.
Rebase duplicatecomparison against already counted commits: the same repository, the same author, the same authoring timestamp and the same set of changed files. File paths are compared, not their content.

The first two rules read only the commit message. The third is the only one that compares a commit against history, and it is the one that catches identifiers rewritten after a rebase or force push, where no annotation is created.

Limits of this mechanism​

They explain results that look like a bug:

  • A revert written by hand will not be recognized. Reverting changes without Git's annotation, for example via git revert --no-commit and a custom message, is, to Q247, an ordinary commit.
  • A cherry-pick without the -x flag leaves no annotation. It will only be caught by the third rule, and only when the author, timestamp and set of files match.
  • A rebase that changes the authoring timestamp (for example via git commit --amend) creates a commit that the third rule will not associate with the original.
  • Only the revert itself is excluded, the reverted commit stays. Work that was reverted is therefore counted once, not zero times.
  • Two different commits by the same author, with the same timestamp and the same set of files, will be treated as a duplicate. This happens with script-generated commits or mass formatting.
  • A manual decision takes precedence. A commit manually included in statistics stays in them regardless of classification, as described in the section below.
Toggling recalculates data gradually

Changing the toggle does not rewrite all commits at once. Q247 fills them in, in batches, in the background, so on a large organization the statistics change over several successive passes, not immediately after saving.

This toggle saves immediately

There is no save button here and no confirmation prompt. Clicking it immediately changes how statistics are calculated for the whole organization.

Excluding a single commit​

Beyond the automatic rule, a specific commit can be excluded. This action takes place outside Configuration: in the commit details window, opened from the increment list on the Project, Team and Participant pages. The window has a two-state toggle, "Exclude from statistics" and "Include in statistics", so the decision can be reversed at any time.

Permissions for this action differ from the rest of this tab. A commit can be excluded by the project manager to whom the repository belongs, or by the team manager. The Admin role alone is not enough: it does not open access to business data, so an administrator without a project assignment will not see this toggle. The full breakdown is described in Permissions.

The exclusion applies to a single commit and covers all metrics calculated from its content: Calories, Increments and Lines and everything derived from them.

See also​