fix: Ignore all changed files with an extension.

There are non-crate files that as far as we know all have file
extensions, as opposed to the crate files we are interested in, which do
not.

Thus skipping all files with extension helps us to get past the initial
commit which includes such files, like `.github/*.yml`.

Related to rust-lang/docs.rs#1807 (comment)

Author: Byron

src/index/diff/delegate.rs

@@ -1,5 +1,6 @@
 use crate::index::diff::Error;
 use crate::{Change, CrateVersion};
+use bstr::BStr;
 use git_repository as git;
 use git_repository::diff::tree::visit::Action;
 use similar::ChangeTag;
@@ -24,6 +25,7 @@ impl<'repo> Delegate<'repo> {
         }
     }
     fn handle(&mut self, change: git::diff::tree::visit::Change) -> Result<(), Error> {
+        use git::bstr::ByteSlice;
         use git::diff::tree::visit::Change::*;
         use git::objs::tree::EntryMode::*;
         fn entry_data(
@@ -36,12 +38,14 @@ impl<'repo> Delegate<'repo> {
                 .transpose()
                 .map_err(Into::into)
         }
-        use git::bstr::ByteSlice;
+        if self.file_name.contains(&b'.') {
+            return Ok(());
+        }
         match change {
             Addition { entry_mode, oid } => {
                 if let Some(obj) = entry_data(self.repo, entry_mode, oid)? {
                     for line in (&obj.data).lines() {
-                        let version = version_from_json_line(line)?;
+                        let version = version_from_json_line(line, self.file_name.as_ref())?;
                         self.changes.push(if version.yanked {
                             Change::Yanked(version)
                         } else {
@@ -70,7 +74,10 @@ impl<'repo> Delegate<'repo> {
                     for change in diff.iter_all_changes() {
                         match change.tag() {
                             ChangeTag::Delete | ChangeTag::Insert => {
-                                let version = version_from_json_line(change.value())?;
+                                let version = version_from_json_line(
+                                    change.value(),
+                                    self.file_name.as_ref(),
+                                )?;
                                 if change.tag() == ChangeTag::Insert {
                                     self.changes.push(if version.yanked {
                                         Change::Yanked(version)
@@ -129,9 +136,10 @@ impl git::diff::tree::Visit for Delegate<'_> {
     }
 }
 
-fn version_from_json_line(line: &[u8]) -> Result<CrateVersion, Error> {
+fn version_from_json_line(line: &[u8], file_name: &BStr) -> Result<CrateVersion, Error> {
     serde_json::from_slice(line).map_err(|err| Error::VersionDecode {
         source: err,
+        file_name: file_name.into(),
         line: line.into(),
     })
 }

src/index/diff/mod.rs

@@ -23,9 +23,10 @@ pub enum Error {
     PeelToTree(#[from] git::object::peel::to_kind::Error),
     #[error("Failed to diff two trees to find changed crates")]
     Diff(#[from] git::diff::tree::changes::Error),
-    #[error("Failed to decode {line:?} as crate version")]
+    #[error("Failed to decode {line:?} in file {file_name:?} as crate version")]
     VersionDecode {
         source: serde_json::Error,
+        file_name: bstr::BString,
         line: bstr::BString,
     },
 }

tests/fixtures/generated-archives/make-index-from-parts.tar.xz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a4c725b07eae662b3f2fcd7a2ae6440fbaaaca0f1a73c9d44fa92f48d34c4a13
-size 582732
+oid sha256:4a5cca6f276fb9ba92fa7f33c645e1daf8e0361acbf1ecf610e4251253b4aa59
+size 586144

tests/fixtures/index-parts/init.9ceec3bd05e9d6ca5a70084cc0078a2f324b66af.tar

@@ -10,17 +10,17 @@ out="$root/index-parts"
 mkdir -p "$out"
 (
   cd "${1:?first argument is the clone of https://github.com/arlosi/crates.io-index}"
-  path=gi/
+  paths="gi/ .github"
   revlist="$root/.tmp.revs"
-  { git log --format=format:%H $path; echo; } | tail -r > "$revlist"
+  { git log --format=format:%H $paths; echo; } | tail -r > "$revlist"
 
   first_commit="$(head -1 "$revlist")"
-  git archive --format tar "$first_commit" $path > "$out/init.$first_commit.tar"
+  git archive --format tar "$first_commit" $paths > "$out/init.$first_commit.tar"
 
   commit_list=$out/commit.list
   tail +2 "$revlist" > "$commit_list"
   while read -r commit; do
-    git diff "$commit"~1.."$commit" -- $path > "$out/$commit".diff
+    git diff "$commit"~1.."$commit" -- $paths > "$out/$commit".diff
     git log --format=%B -n1 "$commit" > "$out/$commit.msg"
   done < "$commit_list"