Differences between revisions 21 and 22
Revision 21 as of 2015-06-04 23:08:30
Size: 8992
Editor: AugieFackler
Revision 22 as of 2015-06-10 20:02:21
Size: 9064
Editor: AugieFackler
Deletions are marked like this. Additions are marked like this.
Line 113: Line 113:
Code is available at [[http://hg.durin42.com/hg-topic-experiment/]].

/!\ This page is primarily intended for Mercurial's developers.

Topic Plan

A (speculative) plan for topic branching that would work more seamlessly with common Mercurial workflows. Still very early prototype stage. Everything is subject to change.

1. Problem Statement

The Mercurial community has been struggling for years to define a nice way to handle 'topic' branches (sometimes also called 'feature' branches), especially when it comes to sharing them with other people (mainly for code review or other collaboration.)

Bookmarks are a clone of git's refs, which seems to work more poorly in Mercurial than they do in Git, in part because the synchronization parts of bookmarks aren't really done. Adding the remaining bits of git's refs to Mercurial has been controversial, and may represent enough of a behavior change that it's infeasible.

Named branches are visible forever in the revision history, which makes them unsuitable for feature branch work as the feature branch names rapidly pollute the output of things like hg branches.

1.1. Goals

The main challenges to get such a feature right are:

- Life cycle:

  • Grouping changes and naming the topic branch is usually only relevant while working to get the feature done. We need a natural/efficient way for the topic to fade away when feature are complete. [augie notes that it might be interesting/useful to be able to query the topic name after the topic is done]

- Distributed:

  • A good solution for topic branch should work well in a distributed environment, where users may be pulling from each other in arbitrary ways. Topic information must be exchanged at the same time as the commits they reference.

- Clearly defined set of changes:

  • A topic is usually composed of multiple changesets. Being able to easily define what is in the set is important for commands that handle topics as a whole (eg histedit and email.) Having a good mechanism for this will probably help produce a better UI in evolution as well. Having a defined set of changesets is also needed when only part of the topic is pushed or pulled. A topic for a change can't be derived from its topological branch and a bookmark because sometimes topics share a root, and sometimes exploration leads to multiple heads on a topic. [TODO(marmoute: work with augie to explain this sentence) It can also be confusing when a topic is rebased.]

- Anonymous branching:

  • Anonymous branches is a useful feature of Mercurial that adds flexibility and improves productivity. We should keep this strength available within each topic.

- name conflict handling:

  • Fixing a bad name (or people fighting over a name) should not result in a very complicated situation (particularly around divergence.)

1.2. Goals Under Debate

- Tracking/Target:

  • People typically make changes with the goal of getting them integrated into a specific line of development (default, stable, version 4.2, staging, etc). This "target" should be the default destination for merge, rebase, update, etc. This aspect can probably use (or maybe need) integration with the life cycle.
    • augie notes that he doesn't like the conceptual complexity this introduces

1.3. Current shortcomings of bookmarks

This plan is not meant as "bookmarks are doomed, lets do something else." The idea here is more driven by "we have some long standing issues with bookmarks, lets think again from scratch and see what emerges." We'll reconcile topics with bookmarks at a later date, once we know what we want. It's entirely possible that bookmarks can grow bits of new functionality and become topics.

There is currently an experimentation around the idea of "remote bookmarks" [TODO(augie): link to a remote bookmarks wiki page] and wider workflow changes, it will be "evaluated" independently. (it's a wiki, please feel free to update the content of this section)




Life cycle






defined set



partial exch












The lifecycle of bookmarks is problematic because we still have not figured out a good way to handle deletion and renaming. Once in the wild, it is very hard to get rid of a feature related bookmark. [augie and marmoute are not sure if remote-bookmarks improves that.]

Bookmark behavior in the distributed area is okay-ish for the in-core bookmark. The current behavior on pull works in a distributed way: it will exchange any bookmark along with the matching changeset. Behavior on push is problematic as the local bookmark may not be pushed (Mercurial may warn about it, but this is not reliable.)

[TODO(marmoute): work with augie to clarify this sentence] The 'remote-boomarks' change is more problematic in this aspect as remote name are not propagated so changesets can get exchanged without there topic information.

tracking is not covered at all by current core version. It is introduced with 'remote-bookmark' but only covers rebase. Additionally, the UI to configure and observe tracking is unclear to [marmoute] yet.

A bookmark can implicitly define a set of revisions since everything 'only' under that bookmark it can be considered in the topic. This has issues:

  • Requires the use of bookmarks for the main branches
  • Misbehaves if a some part of the topic is shared with another bookmark
  • No way of handling of extra anonymous heads on the same topic

Because they refer to a single changeset at the top of the stack, bookmark are bad at partial exchange. It is often practical to push or to pull only a part of the topic because the rest is not ready yet. Because the bookmark have no "start" the shared changesets are pulled anonymously in this case.

For the same reason (referring to a single revision), bookmarks do not allow for experimental branches within a topic label.

Divergent bookmarks provide a solution for conflict. However it does not handle rewinding a bookmark or deletion/recreation cycles.

2. Open ideas

This is a list of idea that emerged while brainstorming.

  • Topic could be a property attached to each changeset (grouping them by similar topic)
  • Topic could fade away when changesets become public (either archived or plain dropped)
    • A benefit of archiving them is that users can query for topics, eg you could say hg log -r topic(issue123) which would help

  • Tracking could be achieved through the naming scheme. eg:
    • 'default//feature-foo' would be a topic 'feature-foo' tracking the 'default' branch.
    • 'stable//issue4700' would be a topic 'issue4700' tracking branch stable.
    • '@//feature-bar' would be a topic 'feature-bar' tracking bookmark '@' ?
    • 'stable//issue4689//issue4700' would be a topic 'issue4700' tracking the topic 'stable//issue4689'. When topic 'issue4686' face away (because published), the tracking fallback to 'stable'.
  • Topics could be non contiguous (mpm idea) feature-foo -> fix-bar -> feature-foo. Allowing a streamlined work that is automatically split apart after that.

  • Topics could be hierarchical 'issue4700.test' 'issue4700.preparation', activation//reference could be done at any level 'issue4700' or 'issue4700' (this could help handle branching/different approach)
  • pushing a new head on a new topic to a non-publishing server would be allowed.

    • that is, it'd be legal to have one head per topic on a non-publishing server.
  • A changeset could maybe have multiple topic.
    • Augie doesn't feel great about this option just because of UI complexity.
  • Users can name patches (in a sense) without mq
    • One of the major complaints about evolve from veteran mq users is that their patches no longer have explicit names. Topics provide a potential way to name patches again.

3. Current Implementation

Assign topics to non-public changesets. A topic is like a named branch, in that it is a label stored in a changeset's extra, but that topics just disappear when the change moves to public phase (the data still exists, it's just not shown.)

Code is available at http://hg.durin42.com/hg-topic-experiment/.

3.0.1. Non-Goals

  • Topics are not suitable for long term branches. We have named branches for that (and possibly also bookmarks, depending on workflow.)
  • Topics are not suitable for tracking a moving point in public history. This seems to be a perfect fit for bookmarks.

3.1. Open Questions

  • Right now we use changeset extra for storing the topic. That might lead to bonus divergence problems. They might be easily fixed, but should we avoid that?
  • Should changesets be allowed multiple topics?
  • How permissive should we be on topic names?

CategoryDeveloper and CategoryNewFeatures

TopicPlan (last edited 2020-02-13 17:58:39 by Pierre-YvesDavid)