On Thu, Nov 05, 2020 at 12:37:49AM +0100, Jakub Narębski wrote: > Hi Abhishek, > > "Abhishek Kumar via GitGitGadget" <gitgitgadget@xxxxxxxxx> writes: > > > This patch series implements the corrected commit date offsets as generation > > number v2, along with other pre-requisites. > > Thanks a lot for continued working on this patch series. Thank you so much for the careful review of the series. > > > > > Git uses topological levels in the commit-graph file for commit-graph > > traversal operations like git log --graph. Unfortunately, using topological > > levels can result in a worse performance than without them when compared > > with committer date as a heuristics. For example, git merge-base v4.8 v4.9 > > on the Linux repository walks 635,579 commits using topological levels and > > walks 167,468 using committer date. > > Very minor nitpick: it would make it easier to read if the commands > themself would be put inside single quotes or backticks, e.g. `git log > --graph` and `git merge-base v4.8 v4.9`. That's unexpected - I wrote the commands within single quotes in the pull request. Since backticks are rendered as "code-tags" on Github, let me try single quotes. > > I wonder if it is worth mentioning (probably not) that this performance > hit was the reason why since 091f4cf3 `git merge-base` uses committer > date heuristics unless there is a cutoff and using topological levels > (generation date v1) is expected to give better performance. > I think that's useful context for someone wondering whether we continue to take the performance hit with topological levels or have abandoned topological levels or chosen some another alternative altogether. > > > > Thus, the need for generation number v2 was born. New generation number > > needed to provide good performance, increment updates, and backward > > compatibility. Due to an unfortunate problem 1 > > Minor issue: this looks a bit strange; is there an error in formatting > this part? Yes. The plaintext in pull request description reads as follows: Thus, the need for generation number v2 was born. New generation number needed to provide good performance, increment updates, and backward compatibility. Due to an unfortunate problem [1], we also needed a way to distinguish between the old and new generation number without incrementing graph version. [1]: https://public-inbox.org/git/87a7gdspo4.fsf@xxxxxxxxxxxxxxxxxxx/ I have been reviewing other pull request descriptions to match their style (and hope the cover letter renders correctly) and Dr. Stolee in his patch series to "add --literal value" option to configuration has written: As reported [1], 'git maintenance unregister' fails when a repository is located in a directory with regex glob characters. [1] https://lore.kernel.org/git/2c2db228-069a-947d-8446-89f4d3f6181a@xxxxxxxxx/T/#mb96fa4187a0d6aeda097cd95804a8aafc0273022 (Note the lack of colon after [1]) > > > [https://public-inbox.org/git/87a7gdspo4.fsf@xxxxxxxxxxxxxxxxxxx/], we also > > needed a way to distinguish between the old and new generation number > > without incrementing graph version. > > > > Various candidates were examined (https://github.com/derrickstolee/gen-test, > > https://github.com/abhishekkumar2718/git/pull/1). The proposed generation > > number v2, Corrected Commit Date with Mononotically Increasing Offsets > > performed much worse than committer date (506,577 vs. 167,468 commits walked > > for git merge-base v4.8 v4.9) and was dropped. > > > > Using Generation Data chunk (GDAT) relieves the requirement of backward > > compatibility as we would continue to store topological levels in Commit > > Data (CDAT) chunk. > > Nice writeup about the history of generation number v2, much appreciated. > > > Thus, Corrected Commit Date was chosen as generation > > number v2. The Corrected Commit Date is defined as: > > Minor nitpick: it would be probably better to use "is defined as > follows." instead of "is defined as:". > > > > > For a commit C, let its corrected commit date be the maximum of the commit > > date of C and the corrected commit dates of its parents plus 1. Then > > corrected commit date offset is the difference between corrected commit date > > of C and commit date of C. As a special case, a root commit with timestamp > > zero has corrected commit date of 1 to be able distinguish it from > > GENERATION_NUMBER_ZERO (that is, an uncomputed corrected commit date). > > Very minor nitpick: s/with timestamp/with *the* timestamp/, and > s/to be able distinguish/to be able *to* distinguish/ (without the '*' > used to mark the additions). > > > > > We will introduce an additional commit-graph chunk, Generation Data chunk, > > Or "Generation DATa chunk", if we want to emphasize where its name came > from, or even "Generation DATa (GDAT) chunk". But it is fine as it is > now, though it would be good idea to write "Generation Data (GDAT) > chunk" to explicitly state its name / shortcut. > > > and store corrected commit date offsets in GDAT chunk while storing > > topological levels in CDAT chunk. The old versions of Git would ignore GDAT > > chunk, using topological levels from CDAT chunk. In contrast, new versions > > of Git would use corrected commit dates, falling back to topological level > > if the generation data chunk is absent in the commit-graph file. > > Nice writeup of handling the backward compatibility. > > > > > While storing corrected commit date offsets saves us 4 bytes per commit (as > > compared with storing corrected commit dates directly), it's possible for > > the offset to overflow the space allocated. To handle such cases, we > > introduce a new chunk, Generation Data Overflow (GDOV) that stores the > > corrected commit date. For overflowing offsets, we set MSB and store the > > position into the GDOV chunk, in a mechanism similar to the Extra Edges list > > chunk. > > Very minor suggestion: perhaps it would be better to use "it's however > possible". > > Very minor suggestion: "it's possible for the offset to overflow" could > be simplified to just "the offset can overflow"... though the simplified > version loses a bit of hint that the overflow should be very rare in > real repositories. > > But it is just fine as it is now; I am not a native English speaker to > judge which version is better. > I think it is better to indicate the rareness of overflows. > > > > For mixed generation number environment (for example new Git on the command > > line, old Git used by GUI client), we can encounter a mixed-chain > > commit-graph (a commit-graph chain where some of split commit-graph files > > have GDAT chunk and others do not). As backward compatibility is one of the > > goals, we can define the following behavior: > > > > While reading a mixed-chain commit-graph version, we fall back on > > topological levels as corrected commit dates and topological levels cannot > > be compared directly. > > > > While writing on top of a split commit-graph, we check if the tip of the > > chain has a GDAT chunk. If it does, we append to the chain, writing GDAT > > chunk. Thus, we guarantee if the topmost split commit-graph file has a GDAT > > chunk, rest of the chain does too. > > > > If the topmost split commit-graph file does not have a GDAT chunk (meaning > > it has been appended by the old Git), we write without GDAT chunk. We do > > write a GDAT chunk when the existing chain does not have GDAT chunk - when > > we are writing to the commit-graph chain with the 'replace' strategy. > > I think the last paragraph can be simplified (or added to) by explicitly > stating the goal: > > When adding new layer to the split commit-graph file, and when merging > some or all layers (replacing them in the latter case), the new layer > will have GDAT chunk if and only if in the final result there would be > no layer without GDAT chunk just below it. > Thanks, that is much clearer to understand. > ... > > After careful review of those 10 patches it looks like the series is > close to being ready, requiring only small changes to progress. > Thank you for writing this handy reference for changes. > > Abhishek Kumar (10): > > commit-graph: fix regression when computing Bloom filters > > All good, beside possible improvement to the commit message. > Thanks to Taylor Blau for discovering possible reason for strange > no change in performance. > > > revision: parse parent in indegree_walk_step() > > Looks good. > > > commit-graph: consolidate fill_commit_graph_info > > Needs to fix now duplicated test names (minor change). > Proposed possible improvement to the commit message. > > > commit-graph: return 64-bit generation number > > Needs fixing due to mismerge: there should be no switch from > using GENERATION_NUMBER_ZERO to using GENERATION_NUMBER_INFINITY. > Possible minor improvement to the commit message. > > > commit-graph: add a slab to store topological levels > > Possible minor improvement to the commit message. > > There is also not very important issue, but something that would be > nice to explain, namely that checks for GENERATION_NUMBER_INFINITY > can never be true, as topo_level_slab_at() returns 0 for commits > outside the commit-graph, not GENERATION_NUMBER_INFINITY. It works > but it is not obvious why. > > > commit-graph: implement corrected commit date > > The change to commit-graph verification needs fixing, and we need to > decide how verifying generation numbers should work. Perhaps a test > for handling topological level of GENERATION_NUMBER_V1_MAX could be > added (though this might be left for ater). > > The changes to `git commit-graph verify` code could be put into > separate patch, either before or after this one. > > > commit-graph: implement generation data chunk > > Proposed possible improvement to the commit message. > The commit message does not explain why given shape of history is > needed to test handling corrected commit date offset overflow. > > Proposed minor corrections to the coding style. > > Instead of looping again through all commits when handling overflow > in corrected commit date offsets, while there should be at most a > few commits needing it, why not save those commits on list and loop > only through those commits? Though this _possible_ performance > improvement could be left to the followup... Since the improvement can be applied to both `write_graph_chunk_generation_data_overflow()` and `write_graph_chunk_extra_edges()`, I am planning to cover this in a followup. > > test_commit_with_date() could be instead implemented via adding > `--date <date>` option to test_commit() in test-lib-functions.sh. > > Also, to reduce "noise" in this patch, the rename of > run_three_modes() to run_all_modes() and test_three_modes() to > test_all_modes() could have been done in a separate preparatory > patch. It would be pure refactoring patch, without introducing any > new functionality. But it is not something that is necessary. > > > commit-graph: use generation v2 only if entire chain does > > Proposed possible improvement to the commit message. > Proposed minor corrections to the coding style (also in tests). > > There is a question whether merging layers or replacing them should > honor GIT_TEST_COMMIT_GRAPH_NO_GDAT. > > Tests possibly could be made more strict, and check more things > explicitly. One test we are missing is testing that merging layers > is done correctly, namely that if we are merging layers in split > commit-graph file, and the layer below the ones we are merging lacks > GDAT chunk, then the result of the merge should also be without GDAT > chunk -- but that might be left for later. > > > commit-reach: use corrected commit dates in paint_down_to_common() > > This patch consist of two slightly interleaved changes, which > possibly could be separated: change to paint_down_to_common() and > change to t6404-recursive-merge test. > > In the commit message for the paint_down_to_common() we should > explicitly mention 091f4cf3, which this one partially reverts. > > Possible accidental change, question about function naming. > > > doc: add corrected commit date info > > Needs further improvements to the documentation, like adding > "[Optional]" to chunk description, and leftover switching from > "generation numbers" to "topological levels" in one place. > > ... > > Best, > -- > Jakub Narębski Thanks - Abhishek
