cli: Add an option to record consecutive snapshots as a single op

CHANGELOG.md

@@ -59,6 +59,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 * Added completions for [Nushell](https://nushell.sh) to `jj util completion`
 
+* Set config `snapshot.squash-consecutive-snapshots = true` to make consecutive
+  snapshots result in a single operation being recorded, effectively
+  automatically abandoning the intermediate snapshot operations, which helps
+  GC, especially if you run `jj log` in a loop for example.
+
 ### Fixed bugs
 
 * On Windows, symlinks in the repo are now materialized as regular files in the
@@ -78,7 +83,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   copy commit on top of a single specified revision, i.e. with one parent.
   `merge` creates a new working copy commit on top of *at least* two specified
   revisions, i.e. with two or more parents.
-  
+
   The only difference between these commands and `jj new`, which *also* creates
   a new working copy commit, is that `new` can create a working copy commit on
   top of any arbitrary number of revisions, so it can handle both the previous
@@ -230,7 +235,7 @@ Thanks to the people who made this release happen!
 
 * `jj branch set` no longer creates a new branch. Use `jj branch create`
   instead.
-  
+
 * `jj init --git` in an existing Git repository now errors and exits rather than
   creating a second Git store.
 
@@ -394,8 +399,8 @@ Thanks to the people who made this release happen!
 
 ### New features
 
-* The `ancestors()` revset function now takes an optional `depth` argument 
-  to limit the depth of the ancestor set. For example, use `jj log -r 
+* The `ancestors()` revset function now takes an optional `depth` argument
+  to limit the depth of the ancestor set. For example, use `jj log -r
   'ancestors(@, 5)` to view the last 5 commits.
 
 * Support for the Watchman filesystem monitor is now bundled by default. Set
@@ -560,13 +565,13 @@ Thanks to the people who made this release happen!
   respectively.
 
 * `jj log` timestamp format now accepts `.utc()` to convert a timestamp to UTC.
- 
+
 * templates now support additional string methods `.starts_with(x)`, `.ends_with(x)`
   `.remove_prefix(x)`, `.remove_suffix(x)`, and `.substr(start, end)`.
 
 * `jj next` and `jj prev` are added, these allow you to traverse the history
   in a linear style. For people coming from Sapling and `git-branchles`
-  see [#2126](https://github.com/martinvonz/jj/issues/2126) for 
+  see [#2126](https://github.com/martinvonz/jj/issues/2126) for
   further pending improvements.
 
 * `jj diff --stat` has been implemented. It shows a histogram of the changes,

cli/src/cli_util.rs

@@ -1413,10 +1413,18 @@ See https://github.com/martinvonz/jj/blob/main/docs/working-copy.md#stale-workin
             let mut tx =
                 start_repo_transaction(&self.user_repo.repo, &self.settings, &self.string_args);
             let mut_repo = tx.mut_repo();
-            let commit = mut_repo
+
+            let mut commit_builder = mut_repo
                 .rewrite_commit(&self.settings, &wc_commit)
-                .set_tree_id(new_tree_id)
-                .write()?;
+                .set_tree_id(new_tree_id);
+
+            let squash_snapshots = self.settings.squash_consecutive_snapshots();
+            if squash_snapshots {
+                commit_builder =
+                    commit_builder.set_predecessors(wc_commit.predecessor_ids().to_owned())
+            }
+
+            let commit = commit_builder.write()?;
             mut_repo.set_wc_commit(workspace_id, commit.id().clone())?;
 
             // Rebase descendants
@@ -1433,6 +1441,21 @@ See https://github.com/martinvonz/jj/blob/main/docs/working-copy.md#stale-workin
                 print_failed_git_export(ui, &failed_branches)?;
             }
 
+            const TAG: &str = "snapshots";
+
+            let mut snapshots = 1;
+
+            if squash_snapshots {
+                if let [parent_op] = tx.parent_ops() {
+                    if let Some(tag) = parent_op.store_operation().metadata.tags.get(TAG) {
+                        snapshots = tag.parse().unwrap_or(1) + 1;
+                        tx.replace_parent()?;
+                    }
+                }
+            }
+
+            tx.set_tag(TAG.into(), snapshots.to_string());
+
             self.user_repo = ReadonlyUserRepo::new(tx.commit("snapshot working copy"));
         }
         locked_ws.finish(self.user_repo.repo.op_id().clone())?;

cli/src/config-schema.json

@@ -357,6 +357,11 @@
                     ],
                     "description": "New files with a size in bytes above this threshold are not snapshotted, unless the threshold is 0",
                     "default": "1MiB"
+                },
+                "squash-consecutive-snapshots": {
+                    "type": "boolean",
+                    "description": "Should snapshot operation happening after another snapshot operation replace it, effectively abandoning the first snapshot and releasing it's view for garbage collection",
+                    "default": false
                 }
             }
         },

cli/src/operation_templater.rs

@@ -138,6 +138,8 @@ fn build_operation_keyword(
             metadata
                 .tags
                 .iter()
+                .filter(|(k, v)| *k != "snapshots" || *v != "1")
+                .sorted_unstable_by_key(|(k, _)| k.to_owned())
                 .map(|(key, value)| format!("{key}: {value}"))
                 .join("\n")
         })),

cli/tests/test_concurrent_operations.rs

@@ -69,10 +69,10 @@ fn test_concurrent_operations_auto_rebase() {
     test_env.jj_cmd_ok(&repo_path, &["describe", "-m", "initial"]);
     let stdout = test_env.jj_cmd_success(&repo_path, &["op", "log"]);
     insta::assert_snapshot!(stdout, @r###"
-    @  d5b4f16ef469 [email protected] 2001-02-03 04:05:08.000 +07:00 - 2001-02-03 04:05:08.000 +07:00
+    @  faefa12d4ced [email protected] 2001-02-03 04:05:08.000 +07:00 - 2001-02-03 04:05:08.000 +07:00
     │  describe commit 123ed18e4c4c0d77428df41112bc02ffc83fb935
     │  args: jj describe -m initial
-    ◉  e632e64d7fa1 [email protected] 2001-02-03 04:05:08.000 +07:00 - 2001-02-03 04:05:08.000 +07:00
+    ◉  75021f3fc19d [email protected] 2001-02-03 04:05:08.000 +07:00 - 2001-02-03 04:05:08.000 +07:00
     │  snapshot working copy
     │  args: jj describe -m initial
     ◉  6ac4339ad699 [email protected] 2001-02-03 04:05:07.000 +07:00 - 2001-02-03 04:05:07.000 +07:00
@@ -184,16 +184,16 @@ fn test_concurrent_snapshot_wc_reloadable() {
     let template = r#"id ++ "\n" ++ description ++ "\n" ++ tags"#;
     let op_log_stdout = test_env.jj_cmd_success(&repo_path, &["op", "log", "-T", template]);
     insta::assert_snapshot!(op_log_stdout, @r###"
-    @  1578600dd63556a22abef7cf6e7054a7e07468187ba31f79d0aa6a197b17004b7cd3e19d2fab1e6a00f2520b48d41969dbbb562c60d4c4af9436224f7f14ab83
+    @  bc38b5d20b2843d546bf5d40d8e1897be131b2ede73b4de031d3b2c6ac394e454a24f52196687accc8d27eb99aaadd8e1d63a8d30a2cb22efae0e3232492bc86
     │  commit 323b414dd255b51375d7f4392b7b2641ffe4289f
     │  args: jj commit -m 'new child1'
-    ◉  90bb10893e980b606939a1f45f2aadf7de1eef65589ac5cd70e20dc20dfd0073c989b5ba0de70ce79a52d27aab5f5699eba66649b531530be5d13bc12c6bd926
+    ◉  af69a63bdc3c88f12e0b1fbc673bce01eb46087b8c6629efd57cb9e68d72973d928f738f2e51504959dc5f531f68a15321f6c9560bca9862f5e850a872ce0c30
     │  snapshot working copy
     │  args: jj commit -m 'new child1'
-    ◉  6104865e95226d46d8c6f5bf43ab025e67f88da6e27f8d8cc598c6d058e333126380c4cb25ea49c841480efee82ce2c602d87b4d3f53b85b4e704af5e83cbdc9
+    ◉  2af019579b0b19d9a147df4779e587b92061fb5cd18fe9ba6d8a33cbeac205ced27c7f411df78278626993725ace7738da9352addb6699898e252c75f57f6f93
     │  commit 3d918700494a9895696e955b85fa05eb0d314cc6
     │  args: jj commit -m initial
-    ◉  76137fc212ef44c53db04be2010ba0419db1fe30e31289bed7d1d0410bee7c3c93d8fd5f6d1b03d93801a2517c436cc1bc4cc512c740e2d88979e771a6fb3730
+    ◉  c07152a7412fac7a7b5570176a1f8799e878c9ec75c7f69633943e6d59a555b7b778804c5b930ce3528e787c5e221c3874b470abb1ff97a2663a4df21f9e09d2
     │  snapshot working copy
     │  args: jj commit -m initial
     ◉  6ac4339ad6999058dd1806653ec37fc0091c1cc17419c750fddc5e8c1a6a77829e6dd70b3408403fb2c0b9839cf6bfd1c270f980674f7f89d4d78dc54082a8ef

cli/tests/test_util_command.rs

@@ -89,7 +89,7 @@ fn test_gc_operation_log() {
     // Now this doesn't work.
     let stderr = test_env.jj_cmd_failure(&repo_path, &["debug", "operation", &op_to_remove]);
     insta::assert_snapshot!(stderr, @r###"
-    Error: No operation ID matching "f9400b5274c6f1cfa23afbc1956349897a6975116135a59ab19d941119cc9fad93d9668b8c7d913fbd68c543dcba40a8d44135a53996a9e8ea92d4b78ef52cb6"
+    Error: No operation ID matching "f642b8a41a8c8da5252704ea63464f01a08ee5d478b6bff26d113b9a53d5e6b632177554eb04a80578447259e0bb4f90a794465991db172290bbc54e75d0d370"
     "###);
 }
 

cli/tests/test_working_copy.rs

@@ -30,3 +30,45 @@ fn test_snapshot_large_file() {
     want this file to be snapshotted. Otherwise add it to your `.gitignore` file.
     "###);
 }
+
+#[test]
+fn test_consecutive_snapshots() {
+    let test_env = TestEnvironment::default();
+    test_env.jj_cmd_ok(test_env.env_root(), &["init", "repo", "--git"]);
+    let repo_path = test_env.env_root().join("repo");
+
+    test_env.add_config(r#"snapshot.squash-consecutive-snapshots = true"#);
+
+    test_env.jj_cmd_ok(&repo_path, &["describe", "-m", "sample text"]);
+    // initial WC is a predecessor of a described WC now
+
+    std::fs::write(repo_path.join("a"), "").unwrap();
+    test_env.jj_cmd_success(&repo_path, &["files"]);
+
+    std::fs::write(repo_path.join("b"), "").unwrap();
+    test_env.jj_cmd_success(&repo_path, &["files"]);
+
+    let stdout = test_env.jj_cmd_success(&repo_path, &["operation", "log"]);
+    insta::assert_snapshot!(stdout, @r###"
+    @  70756af40ab9 [email protected] 2001-02-03 04:05:10.000 +07:00 - 2001-02-03 04:05:10.000 +07:00
+    │  snapshot working copy
+    │  args: jj files
+    │  snapshots: 2
+    ◉  4f5b67be313f [email protected] 2001-02-03 04:05:08.000 +07:00 - 2001-02-03 04:05:08.000 +07:00
+    │  describe commit 230dd059e1b059aefc0da06a2e5a7dbf22362f22
+    │  args: jj describe -m 'sample text'
+    ◉  6ac4339ad699 [email protected] 2001-02-03 04:05:07.000 +07:00 - 2001-02-03 04:05:07.000 +07:00
+    │  add workspace 'default'
+    ◉  1b0049c19762 [email protected] 2001-02-03 04:05:07.000 +07:00 - 2001-02-03 04:05:07.000 +07:00
+    │  initialize repo
+    ◉  000000000000 root()
+    "###);
+
+    let stdout = test_env.jj_cmd_success(&repo_path, &["obslog"]);
+    insta::assert_snapshot!(stdout, @r###"
+    @  qpvuntsm [email protected] 2001-02-03 04:05:10.000 +07:00 bf6e46e6
+    │  sample text
+    ◉  qpvuntsm hidden [email protected] 2001-02-03 04:05:07.000 +07:00 230dd059
+       (empty) (no description set)
+    "###);
+}

cli/tests/test_workspaces.rs

@@ -297,14 +297,14 @@ fn test_workspaces_conflicting_edits() {
     "###);
     let stderr = test_env.jj_cmd_failure(&secondary_path, &["st"]);
     insta::assert_snapshot!(stderr, @r###"
-    Error: The working copy is stale (not updated since operation d93fe4c5a6d1).
+    Error: The working copy is stale (not updated since operation 279268a28063).
     Hint: Run `jj workspace update-stale` to update it.
     See https://github.com/martinvonz/jj/blob/main/docs/working-copy.md#stale-working-copy for more information.
     "###);
     // Same error on second run, and from another command
     let stderr = test_env.jj_cmd_failure(&secondary_path, &["log"]);
     insta::assert_snapshot!(stderr, @r###"
-    Error: The working copy is stale (not updated since operation d93fe4c5a6d1).
+    Error: The working copy is stale (not updated since operation 279268a28063).
     Hint: Run `jj workspace update-stale` to update it.
     See https://github.com/martinvonz/jj/blob/main/docs/working-copy.md#stale-working-copy for more information.
     "###);
@@ -384,7 +384,7 @@ fn test_workspaces_updated_by_other() {
     "###);
     let stderr = test_env.jj_cmd_failure(&secondary_path, &["st"]);
     insta::assert_snapshot!(stderr, @r###"
-    Error: The working copy is stale (not updated since operation d93fe4c5a6d1).
+    Error: The working copy is stale (not updated since operation 279268a28063).
     Hint: Run `jj workspace update-stale` to update it.
     See https://github.com/martinvonz/jj/blob/main/docs/working-copy.md#stale-working-copy for more information.
     "###);
@@ -436,11 +436,11 @@ fn test_workspaces_current_op_discarded_by_other() {
         ],
     );
     insta::assert_snapshot!(stdout, @r###"
-    @  09350a8134 abandon commit acb4b92517b20aa4ee2f3dc58d7c2373754d0b29a3df310dbabda5813f13c3730d28d6a1b6dd37f3b0c8c5c9adaead5dab242ffe7ecc2e5a6a534fe4c6639f89
-    ◉  d8304661a2 Create initial working-copy commit in workspace secondary
-    ◉  4ae0e83b98 add workspace 'secondary'
-    ◉  00cc5ecf50 new empty commit
-    ◉  e0be7087ba snapshot working copy
+    @  4cbdebdfc0 abandon commit acb4b92517b20aa4ee2f3dc58d7c2373754d0b29a3df310dbabda5813f13c3730d28d6a1b6dd37f3b0c8c5c9adaead5dab242ffe7ecc2e5a6a534fe4c6639f89
+    ◉  b4b07f3859 Create initial working-copy commit in workspace secondary
+    ◉  68faf85152 add workspace 'secondary'
+    ◉  ab73dcd486 new empty commit
+    ◉  d47599bc7f snapshot working copy
     ◉  d0bd64e0b3 add workspace 'default'
     ◉  06a74719e6 initialize repo
     ◉  0000000000
@@ -466,7 +466,7 @@ fn test_workspaces_current_op_discarded_by_other() {
 
     let (stdout, stderr) = test_env.jj_cmd_ok(&secondary_path, &["workspace", "update-stale"]);
     insta::assert_snapshot!(stderr, @r###"
-    Failed to read working copy's current operation; attempting recovery. Error message from read attempt: Object d8304661a23b0d8b9ecc517a465869d7c8b6563b460f029541fbe5246cc02718145c5ad8f7fed26863b46a68e79d609b878a7ea5a239baf530858e86e81e72d1 of type operation not found
+    Failed to read working copy's current operation; attempting recovery. Error message from read attempt: Object b4b07f38593831f5a539a17138b9673936dd936f83bf22349c88127b5633cb3d22f94903518a7b673addafe31cc30b3518c30b01e6387417e68594df507cf4b5 of type operation not found
     Created and checked out recovery commit 9d040f9a433c
     "###);
     insta::assert_snapshot!(stdout, @"");
@@ -658,7 +658,7 @@ fn test_workspaces_forget_multi_transaction() {
     // the op log should have multiple workspaces forgotten in a single tx
     let stdout = test_env.jj_cmd_success(&main_path, &["op", "log", "--limit", "1"]);
     insta::assert_snapshot!(stdout, @r###"
-    @  ea093f0a1a06 [email protected] 2001-02-03 04:05:12.000 +07:00 - 2001-02-03 04:05:12.000 +07:00
+    @  119435f7fd1d [email protected] 2001-02-03 04:05:12.000 +07:00 - 2001-02-03 04:05:12.000 +07:00
     │  forget workspaces second, third
     │  args: jj workspace forget second third
     "###);

lib/src/settings.rs

@@ -261,6 +261,12 @@ impl UserSettings {
         }
     }
 
+    pub fn squash_consecutive_snapshots(&self) -> bool {
+        self.config
+            .get_bool("snapshot.squash-consecutive-snapshots")
+            .unwrap_or(false)
+    }
+
     // separate from sign_settings as those two are needed in pretty different
     // places
     pub fn signing_backend(&self) -> Option<String> {

lib/src/transaction.rs

@@ -20,7 +20,7 @@ use itertools::Itertools as _;
 
 use crate::backend::Timestamp;
 use crate::index::ReadonlyIndex;
-use crate::op_store::OperationMetadata;
+use crate::op_store::{OperationId, OperationMetadata};
 use crate::operation::Operation;
 use crate::repo::{MutableRepo, ReadonlyRepo, Repo, RepoLoader, RepoLoaderError};
 use crate::settings::UserSettings;
@@ -32,25 +32,32 @@ pub struct Transaction {
     parent_ops: Vec<Operation>,
     op_metadata: OperationMetadata,
     end_time: Option<Timestamp>,
+    old_ops: Vec<OperationId>,
 }
 
 impl Transaction {
     pub fn new(mut_repo: MutableRepo, user_settings: &UserSettings) -> Transaction {
         let parent_ops = vec![mut_repo.base_repo().operation().clone()];
         let op_metadata = create_op_metadata(user_settings, "".to_string());
         let end_time = user_settings.operation_timestamp();
+        let old_ops = parent_ops.iter().map(|op| op.id().clone()).collect();
         Transaction {
             mut_repo,
             parent_ops,
             op_metadata,
             end_time,
+            old_ops,
         }
     }
 
     pub fn base_repo(&self) -> &Arc<ReadonlyRepo> {
         self.mut_repo.base_repo()
     }
 
+    pub fn parent_ops(&self) -> &[Operation] {
+        &self.parent_ops
+    }
+
     pub fn set_tag(&mut self, key: String, value: String) {
         self.op_metadata.tags.insert(key, value);
     }
@@ -74,12 +81,25 @@ impl Transaction {
         let repo_loader = self.base_repo().loader();
         let base_repo = repo_loader.load_at(&ancestor_op)?;
         let other_repo = repo_loader.load_at(&other_op)?;
+        self.old_ops.push(other_op.id().clone());
         self.parent_ops.push(other_op);
         let merged_repo = self.mut_repo();
         merged_repo.merge(&base_repo, &other_repo);
         Ok(())
     }
 
+    /// Replaces the parent ops with grandparent ops, effectively rewriting the
+    /// parent with self.
+    ///
+    /// Panics if self does not have exactly one parent.
+    pub fn replace_parent(&mut self) -> Result<(), RepoLoaderError> {
+        let [parent]: [Operation; 1] = std::mem::take(&mut self.parent_ops)
+            .try_into()
+            .expect("Merge or root operations cannot replace their parent");
+        self.parent_ops = parent.parents().collect::<Result<_, _>>()?;
+        Ok(())
+    }
+
     /// Writes the transaction to the operation store and publishes it.
     pub fn commit(self, description: impl Into<String>) -> Arc<ReadonlyRepo> {
         self.write(description).publish()
@@ -117,7 +137,7 @@ impl Transaction {
             .index_store()
             .write_index(mut_index, operation.id())
             .unwrap();
-        UnpublishedOperation::new(base_repo.loader(), operation, view, index)
+        UnpublishedOperation::new(base_repo.loader(), operation, view, index, self.old_ops)
     }
 }
 
@@ -148,6 +168,7 @@ pub struct UnpublishedOperation {
     repo_loader: RepoLoader,
     data: Option<NewRepoData>,
     closed: bool,
+    old_ops: Vec<OperationId>,
 }
 
 impl UnpublishedOperation {
@@ -156,6 +177,7 @@ impl UnpublishedOperation {
         operation: Operation,
         view: View,
         index: Box<dyn ReadonlyIndex>,
+        old_ops: Vec<OperationId>,
     ) -> Self {
         let data = Some(NewRepoData {
             operation,
@@ -166,6 +188,7 @@ impl UnpublishedOperation {
             repo_loader,
             data,
             closed: false,
+            old_ops,
         }
     }
 
@@ -179,7 +202,7 @@ impl UnpublishedOperation {
             let _lock = self.repo_loader.op_heads_store().lock();
             self.repo_loader
                 .op_heads_store()
-                .update_op_heads(data.operation.parent_ids(), data.operation.id());
+                .update_op_heads(&self.old_ops, data.operation.id());
         }
         let repo = self
             .repo_loader