Add squash-merging flag to transactions, default to False (#313)

lakeFS v1.48.0 added squash-merging, which, as the name suggests, merges
a source branch into a target branch by squashing the additional history
into a single commit (which is also actually the merge commit in lakeFS).

The interesting thing is that it enables squashing _by default_, which changes
the user-facing behavior of our (merge-based) transactions, and also all
other merges done via `tx.merge()`.

This broke some of our transaction tests, which try to assert correct behavior
by counting commits since branch creation, and verify by commit message.

This commit fixes the tests by introducing a new `squash` flag to the transaction
class, which controls squash-merging of the transaction branch and any other
branches/refs merged via `tx.merge()`.

Interestingly, no special logic is needed to backport the extra keyword argument
to `Reference.merge_into()` - I suspect that this is some pydantic compatibility
detail.

Author: nicholasjng

src/lakefs_spec/transaction.py

@@ -52,6 +52,7 @@ def __init__(self, fs: "LakeFSFileSystem"):
         self.base_branch: Branch | None = None
         self.automerge: bool = False
         self.delete: Literal["onsuccess", "always", "never"] = "onsuccess"
+        self.squash: bool = False
         self._ephemeral_branch: Branch | None = None
 
     def __call__(
@@ -61,6 +62,7 @@ def __call__(
         branch_name: str | None = None,
         automerge: bool = True,
         delete: Literal["onsuccess", "always", "never"] = "onsuccess",
+        squash: bool = False,
     ) -> "LakeFSTransaction":
         """
         Creates an ephemeral branch, conducts all uploads and operations on that branch,
@@ -83,6 +85,8 @@ def __call__(
             or failure.
 
             If ``"never"``, the transaction branch is always left in the repository.
+        squash: bool
+            Optionally squash-merges the transaction branch into the base branch.
         """
 
         if isinstance(repository, str):
@@ -102,6 +106,7 @@ def __call__(
 
         self.automerge = automerge
         self.delete = delete
+        self.squash = squash
 
         ephem_name = branch_name or "transaction-" + "".join(random.choices(string.digits, k=6))  # nosec: B311
         self._ephemeral_branch = Branch(self.repository, ephem_name, client=self.fs.client)
@@ -136,7 +141,7 @@ def __exit__(self, exc_type, exc_val, exc_tb):
 
         if success and self.automerge:
             if any(self.base_branch.diff(self._ephemeral_branch)):
-                self._ephemeral_branch.merge_into(self.base_branch)
+                self._ephemeral_branch.merge_into(self.base_branch, squash_merge=self.squash)
         if self.delete == "always" or (success and self.delete == "onsuccess"):
             self._ephemeral_branch.delete()
 
@@ -170,7 +175,7 @@ def commit(self, message: str, metadata: dict[str, str] | None = None) -> Refere
 
         return self.branch.commit(message, metadata=metadata)
 
-    def merge(self, source_ref: str | Branch, into: str | Branch) -> Commit:
+    def merge(self, source_ref: str | Branch, into: str | Branch, squash: bool = False) -> Commit:
         """
         Merge a branch into another branch in a repository.
 
@@ -184,6 +189,8 @@ def merge(self, source_ref: str | Branch, into: str | Branch) -> Commit:
             Can be a branch name or partial commit SHA.
         into: str | Branch
             Target branch into which the changes will be merged.
+        squash: bool
+            Optionally squash-merges the source reference into the target branch.
 
         Returns
         -------
@@ -194,7 +201,7 @@ def merge(self, source_ref: str | Branch, into: str | Branch) -> Commit:
         dest = _ensurebranch(into, self.repository, self.fs.client)
 
         if any(dest.diff(source)):
-            source.merge_into(dest)
+            source.merge_into(dest, squash_merge=squash)
         return dest.head.get_commit()
 
     def revert(self, branch: str | Branch, ref: ReferenceType, parent_number: int = 1) -> Commit: