dirstate: restore original estimation and update comment
The former comment didn't reflect the content of the dirstate entries,
the two nodes are a fixed header in the file and not per-entry. Split
the documented entry into the path part and the fixed header. The
heuristic itself is still valid, e.g. for the NetBSD src tree a maximum
path size of 142 and an average of 49, resulting in 66 bytes per entry
on average.
Differential Revision: https://phab.mercurial-scm.org/D8850
--- a/mercurial/dirstate.py Fri Jul 31 17:09:31 2020 +0530
+++ b/mercurial/dirstate.py Thu Jul 30 16:13:17 2020 +0200
@@ -1658,18 +1658,12 @@
if util.safehasattr(parsers, b'dict_new_presized'):
# Make an estimate of the number of files in the dirstate based on
- # its size. From a linear regression on a set of real-world repos,
- # all over 10,000 files, the size of a dirstate entry is 2 nodes
- # plus 45 bytes. The cost of resizing is significantly higher than the cost
- # of filling in a larger presized dict, so subtract 20% from the
- # size.
- #
- # This heuristic is imperfect in many ways, so in a future dirstate
- # format update it makes sense to just record the number of entries
- # on write.
- self._map = parsers.dict_new_presized(
- len(st) // ((2 * self._nodelen + 45) * 4 // 5)
- )
+ # its size. This trades wasting some memory for avoiding costly
+ # resizes. Each entry have a prefix of 17 bytes followed by one or
+ # two path names. Studies on various large-scale real-world repositories
+ # found 54 bytes a reasonable upper limit for the average path names.
+ # Copy entries are ignored for the sake of this estimate.
+ self._map = parsers.dict_new_presized(len(st) // 71)
# Python's garbage collector triggers a GC each time a certain number
# of container objects (the number being defined by