D7889: nodemap: track the total and unused amount of data in the rawdata file

marmoute (Pierre-Yves David) phabricator at mercurial-scm.org
Wed Jan 15 14:58:27 UTC 2020


marmoute created this revision.
Herald added a subscriber: mercurial-devel.
Herald added a reviewer: hg-reviewers.

REVISION SUMMARY
  We need to keep that information around:
  
  - total data will allow transaction to start appending new information without confusing other reader.
  
  - unused data will allow to detect when we should regenerate new rawdata file.

REPOSITORY
  rHG Mercurial

REVISION DETAIL
  https://phab.mercurial-scm.org/D7889

AFFECTED FILES
  mercurial/debugcommands.py
  mercurial/pure/parsers.py
  mercurial/revlogutils/nodemap.py
  tests/test-persistent-nodemap.t

CHANGE DETAILS

diff --git a/tests/test-persistent-nodemap.t b/tests/test-persistent-nodemap.t
--- a/tests/test-persistent-nodemap.t
+++ b/tests/test-persistent-nodemap.t
@@ -15,8 +15,10 @@
   $ hg debugnodemap --metadata
   uid: ???????????????? (glob)
   tip-rev: 5000
+  data-length: 245760
+  data-unused: 0
   $ f --size .hg/store/00changelog.n
-  .hg/store/00changelog.n: size=26
+  .hg/store/00changelog.n: size=42
   $ f --sha256 .hg/store/00changelog-*.nd
   .hg/store/00changelog-????????????????.nd: sha256=bc400bf49f11e83bbd25630439feee6628a80a8602d2e38972eac44cc3efe10c (glob)
   $ hg debugnodemap --dump-new | f --sha256 --size
@@ -50,11 +52,22 @@
   $ echo foo > foo
   $ hg add foo
   $ hg ci -m 'foo'
+
+#if pure
   $ hg debugnodemap --metadata
   uid: ???????????????? (glob)
   tip-rev: 5001
+  data-length: 246144
+  data-unused: 384
+#else
+  $ hg debugnodemap --metadata
+  uid: ???????????????? (glob)
+  tip-rev: 5001
+  data-length: 245760
+  data-unused: 0
+#endif
   $ f --size .hg/store/00changelog.n
-  .hg/store/00changelog.n: size=26
+  .hg/store/00changelog.n: size=42
 
 (The pure code use the debug code that perform incremental update, the C code reencode from scratch)
 
diff --git a/mercurial/revlogutils/nodemap.py b/mercurial/revlogutils/nodemap.py
--- a/mercurial/revlogutils/nodemap.py
+++ b/mercurial/revlogutils/nodemap.py
@@ -37,10 +37,12 @@
         return None
     offset += S_VERSION.size
     headers = S_HEADER.unpack(pdata[offset : offset + S_HEADER.size])
-    uid_size, tip_rev = headers
+    uid_size, tip_rev, data_length, data_unused = headers
     offset += S_HEADER.size
     docket = NodeMapDocket(pdata[offset : offset + uid_size])
     docket.tip_rev = tip_rev
+    docket.data_length = data_length
+    docket.data_unused = data_unused
 
     filename = _rawdata_filepath(revlog, docket)
     return docket, revlog.opener.tryread(filename)
@@ -78,12 +80,14 @@
     # first attemp an incremental update of the data
     if can_incremental and ondisk_docket is not None:
         target_docket = revlog._nodemap_docket.copy()
-        data = revlog.index.nodemap_data_incremental()
+        data_changed_count, data = revlog.index.nodemap_data_incremental()
         datafile = _rawdata_filepath(revlog, target_docket)
         # EXP-TODO: if this is a cache, this should use a cache vfs, not a
         # store vfs
         with revlog.opener(datafile, 'a') as fd:
             fd.write(data)
+        target_docket.data_length += len(data)
+        target_docket.data_unused += data_changed_count
     else:
         # otherwise fallback to a full new export
         target_docket = NodeMapDocket()
@@ -96,6 +100,7 @@
         # store vfs
         with revlog.opener(datafile, 'w') as fd:
             fd.write(data)
+        target_docket.data_length = len(data)
     target_docket.tip_rev = revlog.tiprev()
     # EXP-TODO: if this is a cache, this should use a cache vfs, not a
     # store vfs
@@ -143,9 +148,8 @@
 
 # version 0 is experimental, no BC garantee, do no use outside of tests.
 ONDISK_VERSION = 0
-
 S_VERSION = struct.Struct(">B")
-S_HEADER = struct.Struct(">BQ")
+S_HEADER = struct.Struct(">BQQQ")
 
 ID_SIZE = 8
 
@@ -168,17 +172,26 @@
             uid = _make_uid()
         self.uid = uid
         self.tip_rev = None
+        self.data_length = None
+        self.data_unused = 0
 
     def copy(self):
         new = NodeMapDocket(uid=self.uid)
         new.tip_rev = self.tip_rev
+        new.data_length = self.data_length
+        new.data_unused = self.data_unused
         return new
 
     def serialize(self):
         """return serialized bytes for a docket using the passed uid"""
         data = []
         data.append(S_VERSION.pack(ONDISK_VERSION))
-        headers = (len(self.uid), self.tip_rev)
+        headers = (
+            len(self.uid),
+            self.tip_rev,
+            self.data_length,
+            self.data_unused,
+        )
         data.append(S_HEADER.pack(*headers))
         data.append(self.uid)
         return b''.join(data)
@@ -236,8 +249,8 @@
 def update_persistent_data(index, root, max_idx, last_rev):
     """return the serialised data of a nodemap for a given index
     """
-    trie = _update_trie(index, root, last_rev)
-    return _dump_trie(trie, existing_idx=max_idx)
+    changed_block, trie = _update_trie(index, root, last_rev)
+    return changed_block * S_BLOCK.size, _dump_trie(trie, existing_idx=max_idx)
 
 
 S_BLOCK = struct.Struct(">" + ("q" * 16))
@@ -293,10 +306,11 @@
 
 def _update_trie(index, root, last_rev):
     """consume"""
+    changed = 0
     for rev in range(last_rev + 1, len(index)):
         hex = nodemod.hex(index[rev][7])
-        _insert_into_block(index, 0, root, rev, hex)
-    return root
+        changed += _insert_into_block(index, 0, root, rev, hex)
+    return changed, root
 
 
 def _insert_into_block(index, level, block, current_rev, current_hex):
@@ -308,6 +322,7 @@
     current_rev: the revision number we are adding
     current_hex: the hexadecimal representation of the of that revision
     """
+    changed = 1
     if block.ondisk_id is not None:
         block.ondisk_id = None
     entry = block.get(_to_int(current_hex[level]))
@@ -316,7 +331,9 @@
         block[_to_int(current_hex[level])] = current_rev
     elif isinstance(entry, dict):
         # need to recurse to an underlying block
-        _insert_into_block(index, level + 1, entry, current_rev, current_hex)
+        changed += _insert_into_block(
+            index, level + 1, entry, current_rev, current_hex
+        )
     else:
         # collision with a previously unique prefix, inserting new
         # vertices to fit both entry.
@@ -329,6 +346,7 @@
             level += 1
         block[_to_int(current_hex[level])] = current_rev
         block[_to_int(other_hex[level])] = other_rev
+    return changed
 
 
 def _dump_trie(root, existing_idx=None):
diff --git a/mercurial/pure/parsers.py b/mercurial/pure/parsers.py
--- a/mercurial/pure/parsers.py
+++ b/mercurial/pure/parsers.py
@@ -164,11 +164,11 @@
         """
         if self._nm_root is None:
             return None
-        data = nodemaputil.update_persistent_data(
+        changed, data = nodemaputil.update_persistent_data(
             self, self._nm_root, self._nm_max_idx, self._nm_rev
         )
         self._nm_root = self._nm_max_idx = self._nm_rev = None
-        return data
+        return changed, data
 
     def update_nodemap_data(self, docket, nm_data):
         """provide full serialiazed data from a nodemap
diff --git a/mercurial/debugcommands.py b/mercurial/debugcommands.py
--- a/mercurial/debugcommands.py
+++ b/mercurial/debugcommands.py
@@ -2131,6 +2131,8 @@
             docket, data = nm_data
             ui.write((b"uid: %s\n") % docket.uid)
             ui.write((b"tip-rev: %d\n") % docket.tip_rev)
+            ui.write((b"data-length: %s\n") % docket.data_length)
+            ui.write((b"data-unused: %s\n") % docket.data_unused)
 
 
 @command(



To: marmoute, #hg-reviewers
Cc: mercurial-devel


More information about the Mercurial-devel mailing list