stablerange: adds some caching of the subranges The goal of subranges is to be as reusable as possible, so we cache the subrange relationship to take advantage of this.

stablerange: operate on an unfiltered repository We only do operation on ancestors so using unfiltered is fine.

depth: use lower lever function to compute missing The time spend parsing the revset is noticeable.

depth: update depth to code to reuse ancestors depth Computing the depth of all N revs is no longer 'O(N**2)'

obsdiscovery: extract a smarted depth in utility The function is reusing previous depth for ancestors unless this is a merge.

obsdiscovery: simply some of the missing computation

discovery: implement some range based discovery This use the previously implemented methods to implement some discovery and fetch method using ranges. There is currently major performance issue and some aspect of the protocol will changes.

discovery: introduce a official 'obshash associated to a range The hash in a combination of the hash a its subrange. If the range is a leaf, we have the obsmarker relevant to the node in that leaf range.

discovery: introduce "stable slicing" methods We introduce new code that leverage the stable sorting to slices a graph in a way "stable" accross repository. This should allow us to use theses slices for obsolescence markers discovery.

discovery: adds a function doing revnum independant sorting The function is doing a topological sort without depending on revision number. This make the sort independent of the repository and suitable for discovery.

hgignore: also ignore the 'dist' directory This is generated by the 'sdist' command

obsexchange: drop unused module shame shame shame

MANIFEST.in: also includes topic That was also overlooked when we added topic.

MANIFEST.in: fix to include all evolve submodule This was overlooked before.

obsdiscovery: organize the code by category A private function get promoted to public since another module uses it.

evolution: drop the old __temporary__.advertiseobsolete option Mercurial code supports such feature directly though its 'experimental.evolution' config option. So we drop the old code and config.

README: mention the 'serveronly' extensions It is now properly installed with the package.

server only: test a more standard method of disabling exchange We have an "official" config option from core so lets just use it.

debian: don't delete evolve and topic extensions The debian packaging rules deletes all __init__.py to avoid conflicting with the __init__.py that declares the namespace package. Now that the evolve and topic extensions also live in __init__.py files, we clearly don't want to delete all __init__.py files anymore; only the hgext3rd/__init__.py should be deleted.

serveronly: replace exc.message with str(exc) BaseException.message is deprecated in Python 2.6, so use str() instead.

cleanup: stop using 'repo.join' methods These repository method might get deprecated soon. In addition, accessing the vfs-s is not that hard.

setup: add author_email The 'sdist' command complains about it otherwise.

readme: make it valid rst That will eventually please pypi

setup.py: remove unused and harmful code importing mercurial This was failing on some build system we have. The import seemed to succeed, but then mercurial.__all__ would fail. I don't know why, but since it's unused anyway, let's just drop it.

docs: update references to installation path

test: rename serveronly tests to reflect the new extension name

evolve: stop running 'serveronly' setup Since there is no common logic remaining the smaller extension we can stop running its setup in the main extensions

discovery: directly merge the 'obsdiscovery' exthelper in top level extensions There is no reason to proxy it through the 'obsexchange' submodule.

compat: drop advertising of unsupported experimental bundle2 part The support for this part was dropped long ago in 74bc8a0c2c02. But we forgot to remove it from the 'simple4server' extension.

exchange: dispatch all code in 'serveronly' to the appropriate submodule All code related to exchange is now in either 'evolve.obsexchange' or 'evolve.obsdiscovery', and the 'serveronly' extension is using their 'exthelper' to set itself up. As a side effect, the 'serveronly' extensions now enable better exchange when pushing from that server too.

capabilities: properly sort capabilities after updating them Capabilities are usually sorted. This is especially useful to avoid silly error in test.

pull: also issue warning for old mercurial version

exchange: split between modern and legacy code We make it clearer what code is old crust and what code is the cutting edge part.

push: add extra warning about pushing to old server Pushing using old method is slow and racy. We adds warning to point this to the user. We also add inline comment to clarify the function purpose.

push: exit obsexchange early if disabled We were exiting the function later, in particular -after- issuing a message about exchanging obsmarkers.

compat: drop special code handling change in push/pull op API All version of Mercurial we support have the same API.

discovery: split discovery related code in 'obsdiscovery' More code splitting for more clarity.

exchange: rename the module Let's add a 'obs' prefix to make the module role clearer.

setup.py: include all evolve submodule We are splitting more and more code into submodule, we make sure their are properly installed. As a side effect the 'legacy' module is now installed, that is not too important. The 'evolve.hack' package is still non installed by default, as intended.

evolve: switch away from deprecated repo.opener

serveronly: use the same extensions metadata than evolve This will avoid them to drift out of sync with evolve.

evolve: move extension metadata in their own module This will allow the server only extension to reuse them. As a side effect we can now simplify the config parsing in setup.py

evolve: specify the 'minimalhgversion' 3.7 added a new nice and shinny variable to help user know when they use the wrong version of Mercurial with an extension. We now use it.

serveronly: give the sub extension a way to access to the 'evolve' module We keep it as clean as possible but if the extension is specified with a direct path, we have to be a bit hacky. Being able to access the whole evolve extension from the 'serveronly' extension will lift multiple constraints on how we organise the code and will allow for cleaner and clearer code. We extract a minor function into a 'utility' module to have something to depends on.

test: update test to check for various way to import the serveronly things We add a "setup script" that update the PYTHONPATH to allow direct import of the extension code. Then we tests that all way to import the extensions works.

tests: move "test setup" script into a 'testlib' directory This will avoid cluttering the tests directory.

exchange: move code related to exchange into a 'evolve.exchange' submodule The evolve extension is HUGE, we split exchange code appart before doing more work on it.

exthelper: move command declaration into the exthelper This is necessary to allow sub module to declare commands too.

exthelper: add a 'merge' method This will help us spread more code in different sub module.

exthelper: extract into its own submodule That code is fairly independant, we extract it into its own extensions to clarify the code

makefile: add an 'install-home' target It is similar to the Mercurial one.

makefile: only check for HGROOT when running tests Otherwise we breaks other target that do not needs HGROOT.

tests: split pyflake and flake8 tests This allow more parallelism and better skip message

flake8: run on all python file We can now run it on all file in the repository to catch potential regression.

flake8: fix error in 'setup.py' Some strange indentation in there.

flake8: fix error in 'test2rst.py' There is variosu dead code in there :-/ but we just fix flake8 complains for now.

flake8: remove "error" in documentation configuration The "error" are debatable but cheap to fix.

flake8: enable on evolve too Now that all error are fixed we can run it on all file to prevent regression.

flake8: update the code sources to fixes a horde of pyflake warnings There is many of them, they all fixed.

flake8: register 'xrange' as a builtins For some reason, flake8 thinks is is undefined.

flake8: ignore an addition rules related to import Our initialization code is too complex to make flake8 happy

flake8: document all the exception we added

README: mention the addition of the topic extensions That is probably worth noting.

tests: run pyflake on all python files We have fixed the handfull of error pyflakes found on evolve so we can setup the test to prevent regression. Next stop flake8.

serveronly: fix undefined variable This is a legit bug caught by pyflake

legacy: properly test for mercurial version in the legacy extensions Otherwise 'demandimport' would delay the raising of ImportError.

evolve: prevent global variable shadowing Small and harmless pyflake complains.

evolve: drop some unused variables More pyflakes catches.

evolve: drop some unused import topic came with pyflake checking, let us make use of it.

README: fix evolve installation path We forgot to do so when we moved things to hgext3rd.

topic: merge the topic extension in the evolve repository There is a lot of synergy between the two concepts. Topic is expected to be able to smooth multiple of evolution sharp edge. Having them both in the same repository will make this collaboration easier.

test: adapt to new core output Troubles are now displayed by log so we have to update the test.

test: adapt to change in mercurial output Nothing major here.

compat: handle difference of signature for update destination hooks The 'check' argument have been dropped in 11c253997b0e. So we make it optional for compatibility.

setup.py: also install 'hgext3rd/__init__.py' This file is needed for the namespace package to work. (And also for the package to work on its own at all).

tutorial: add a very basic pointer in the readme Ideally we should at least build the html from the rst.

tutorial: add a second part about having multiple topic

tutorial: add a first version of the tutorial with the basics This is probably not wonderful but at least we have a base.

doc: add more details about the head checking Once again, probably not great, but better than nothing.

documentation: some basic update of the documentation So far the documentation was pretty scarce and a lot of important information was missing. The new content is probably not great but it is an improvement.

flake8: ignore N801 N801 is "class names should use CapWords convention" The message comes from a pep8 plugin "pep8-naming" which cares about naming conventions. If the plugin is installed, it is automatically used by pep8 and flake8.

flake8: don't ignore E713 now that it's fixed E713 is "test for membership should be ‘not in’"

flake8: fix E713 style

flake8: fix E301 style

flake8: fix E111 style OFF BY ONE ERRORS ARE THE WORST

flake8: fix E222 style

flake8: fix E127 style

flake8: fix W391 style

tests: add flake8 check This test will hopefully help prevent style errors. To be determined: which errors / warnings to ignore?

flake8: add file for rules to ignore This will help prevent users that have an editor with flake8 capabilities from fixing linter errors for rules that are not wanted by the project. Initially, this ignore list includes everything in the project that violates pep8. I'll try to ween this list down to something a little more manageable.

init: guard _namemap with repo.topics (issue5351) This prevents reading extra data on all repo commits when there is no topic for a given name. _namemap is called a lot, and it is often called for commit shas -- which should pretty much never be names of topics...

init: cache repo.topics

stack: add warning about multiple heads This seems a good spot to nudge people toward reducing there head count.

stack: add some behind information This is the next useful thing after branch

stack: add some basic branch information In the same spirit as knowing the topic you are on, having some branch data seem useful (yet, this smell a bit like summary)

stack: add some header with the topic name This is the first step of a move to add more data about a stack in the command output.

test: add testing of stack output in 'stack-data' test. We are going to display common information, so we reuse test case.

test: rename 'list' test to 'stack-data' We are about to add stack data information to 'hg stack'. As we'll re-use the test data creating the 'test-topic-list.t' we need to rename it first to embrace its new semantic.

stack: also dim bases Now that they may have index it seems better to have a stronger marking for bases.

stack: reusing the index number in base when applicable This clarify the branching when it is easy to track back the branching point. This does not takes the evolution graph into account yet.

stack: extra argument validation logic outside of showstack It seems sensible keep the 'showstack' function purely about actually displaying and existing stack.

summary: properly label the topic name in summary This should be labelled 'topic.active' so that color can pick it up properly.

topic: list the branches this topic belong to when verbose Using the verbose listing a bit more, this seems to be useful information.

topic: properly justify the verbose data when listing topic We align all data to the widest topic.

ngtip: rely on topicmap for 'ngtip' ngtip is an old function that predates the introduction of a more proper topic aware 'topicmap'. We replace the old code with a call to the new one. This change set function return type and have minor effect in a couple of place.

topic: list the number of 'behind' changeset when --verbose is used Displaying more information in the topic list is useful, we continue with the number of 'behind' changesets. This 'behind' count the number of new changesets on the default rebase destination. This will highlight topics that need rebasing.

destination: honor the sourceset argument of _destmerge Mercurial gained a "sourceset" argument to compute the merge destination of any changeset (the one in the 'sourceset'). This now works for the version in topic too. The support is quite basic (because we pick only one changeset from the sourceset) but that's a start.

destmerge: improve topic head computation and handling This rework get use two benefits: - heads are now computed on a "stabilized" state, - heads above the 'source' are now properly cleaned up,

evolve: extract the code copied from evolve in a submodule This will allow us to use that code in other module.

destmerge: honor the 'onheadcheck' argument This argument was introduced in Mercurial some time ago but the topic extension was never updated to take it in account. This is now fixed.

destination: use the correct abort class for error These errors have been in Mercurial long enough. Having them will make the coming changesets simpler.

topic: list the number of head when --verbose is used Displaying more information in the topic list is useful, we continue with the number of heads on the topics, this will help highlight the problem to users. We only display head if there is more than one.

topic: list the number of troubled changesets when --verbose is used Displaying more information in the topic list is useful, we continue with the number of troubled changesets. This will help people to spot topics that needs evolution, for example after a pull.

topic: list the number of changesets when --verbose is used Displaying more information in the topic list is useful, we start with the most obvious: number of non-obsolete changeset in the topic.

topic: add color for the active topic We reuse the same color that the one used for the active bookmark.

topic: add formatter support Using formatter to output the list of topic will add support for color and richer templater. We add some test for output control. We introduce a dedicated test to help with coming more complexe output.

topic: extract the code listing all topics The code will become more complexe with formatter support and additional data in the verbose output. We extract the code in its own function first.

stack: introduce and explicite command to display the stack If is worth testing having the feature lives outside 'hg topic --list'. The 'stack' name is sub optimal because the conflict with 'status'. using "hg st" would still work as 'st' is hard coded, but this mean 'hg stack' has no short version.

testedwith: update to 3.9 This is might still 3.7 compatible but I'm no longer running the test and I would says the user base is small enough to not care.

topic: make sure we have the 'wlock' when setting topic We where writing the '.hg/topic' file without a lot. That was bad.