diffSummary is incorrect for the first commit

[web-dev-sam]

When comparing the first commit with the empty tree to see what changes the first commit introduced it displays the "changes" correctly but additions/deletions are wrong.

const EMPTY_TREE = "4b825dc642cb6eb9a060e54bf8d69288fbee4904"
const fromRepo = simpleGit("...")
console.log(await fromRepo.diffSummary([EMPTY_TREE, await fromRepo.firstCommit()]))

Git Output: git diff 4b825dc642cb6eb9a060e54bf8d69288fbee4904 8a424bd942b041bc924989e8ab3c77a63e26826d

diff --git a/.env.example b/.env.example
new file mode 100644
index 0000000..5d329db
--- /dev/null
+++ b/.env.example
@@ -0,0 +1,12 @@
+NUXT_AUTH_ORIGIN= # No trailing slash
+NUXT_SESSION_PASSWORD=
+
+NUXT_OAUTH_GOOGLE_CLIENT_ID=
+NUXT_OAUTH_GOOGLE_CLIENT_SECRET=
+NUXT_OAUTH_GITHUB_CLIENT_ID=
+NUXT_OAUTH_GITHUB_CLIENT_SECRET=
+NUXT_OAUTH_GITHUB_CONNECT_CLIENT_ID=
+NUXT_OAUTH_GITHUB_CONNECT_CLIENT_SECRET=
+
+NUXT_POSTGRES_URL=
+RESEND_API_KEY=

git-js Output:

{
      file: ".env.example",   // correct
      changes: 12,            // correct
      insertions: 4,          // incorrect
      deletions: 0,           // incorrect
      binary: false           // correct
}

[web-dev-sam]

This is my fix for now:

const EMPTY_TREE = { hash: "4b825dc642cb6eb9a060e54bf8d69288fbee4904" }
const diffs = []
for (let i = all.length - 1; i > 0; i--) {
  const isFirst = i === all.length - 1

  const olderCommit = isFirst ? EMPTY_TREE : all[i]
  const newerCommit = isFirst ? all[i] : all[i-1]
  const diff = await fromRepo.diffSummary([olderCommit.hash, newerCommit.hash])

  // Fix bug where diffSummary doesn't return correct values for the first commit
  if (isFirst) {
    diff.insertions = diff.changed
    diff.deletions = 0
    diff.files = diff.files.map(f => {
      if (f.binary) {
        return f
      }
      f.insertions = f.changes
      f.deletions = 0
      return f
    })
  }
  diffs.push(diff)
}

[steveukx]

Hello, thank you for the detailed issue description!

The cause of the problem is that there are files larger than 4096b in size in the initial commit, so the --stat=4096 used to count the size of the insertion/deletion for each file is being rounded to a percentage rather than showing the full detail.

For a diff between the empty tree commit and any other commit, you can avoid this by using the --numstat mechanism for detecting changes, which will always be the same as having used the (default) --stat mechanism when diffing against the empty tree commit:

await fromRepo.diffSummary(['--numstat', EMPTY_TREE, await fromRepo.firstCommit()]));