view mercurial/help/revsets.txt @ 29861:2f6d5c60f6fc stable

annotate: pre-calculate the "needed" dictionary (issue5360)

The "needed" dict is used as a reference counter to free items in the giant
"hist" dict. However, currently it is not very accurate and can lead to
dropping "hist" items unnecessarily, for example, with the following DAG,

       -3-
      /   \
  0--1--2--4--

The current algorithm will visit and calculate rev 1 twice, undesired. And
it tries to be smart by clearing rev 1's parents: "pcache[1] = []" at the
time hist[1] being accessed (note: hist[1] needs to be used twice, by rev 2
and rev 3). It can result in incorrect results if p1 of rev 4 deletes chunks
belonging to rev 0.

However, simply removing "needed" is not okay, because it will consume 10x
memory:

  # without any change
  % HGRCPATH= lrun ./hg annotate mercurial/commands.py -r d130a38 3>&2 [1]
  MEMORY   49074176
  CPUTIME  9.213
  REALTIME 9.270

  # with "needed" removed
  MEMORY   637673472
  CPUTIME  8.164
  REALTIME 8.249

This patch moves "needed" (and "pcache") calculation to a separate DFS to
address the issue. It improves perf and fixes issue5360 by correctly reusing
hist, while maintaining low memory usage. Some additional attempt has been
made to further reduce memory usage, like changing "pcache[f] = []" to "del
pcache[f]". Therefore the result can be both faster and lower memory usage:

  # with this patch applied
  MEMORY   47575040
  CPUTIME  7.870
  REALTIME 7.926

[1]: lrun is a lightweight sandbox built on Linux cgroup and namespace. It's
     used to measure CPU and memory usage here. Source code is available at
     github.com/quark-zju/lrun.

Mercurial supports a functional language for selecting a set of
revisions.

The language supports a number of predicates which are joined by infix
operators. Parenthesis can be used for grouping.

Identifiers such as branch names may need quoting with single or
double quotes if they contain characters like ``-`` or if they match
one of the predefined predicates.

Special characters can be used in quoted identifiers by escaping them,
e.g., ``\n`` is interpreted as a newline. To prevent them from being
interpreted, strings can be prefixed with ``r``, e.g. ``r'...'``.

There is a single prefix operator:

``not x``
  Changesets not in x. Short form is ``! x``.

These are the supported infix operators:

``x::y``
  A DAG range, meaning all changesets that are descendants of x and
  ancestors of y, including x and y themselves. If the first endpoint
  is left out, this is equivalent to ``ancestors(y)``, if the second
  is left out it is equivalent to ``descendants(x)``.

  An alternative syntax is ``x..y``.

``x:y``
  All changesets with revision numbers between x and y, both
  inclusive. Either endpoint can be left out, they default to 0 and
  tip.

``x and y``
  The intersection of changesets in x and y. Short form is ``x & y``.

``x or y``
  The union of changesets in x and y. There are two alternative short
  forms: ``x | y`` and ``x + y``.

``x - y``
  Changesets in x but not in y.

``x % y``
  Changesets that are ancestors of x but not ancestors of y (i.e. ::x - ::y).
  This is shorthand notation for ``only(x, y)`` (see below). The second
  argument is optional and, if left out, is equivalent to ``only(x)``.

``x^n``
  The nth parent of x, n == 0, 1, or 2.
  For n == 0, x; for n == 1, the first parent of each changeset in x;
  for n == 2, the second parent of changeset in x.

``x~n``
  The nth first ancestor of x; ``x~0`` is x; ``x~3`` is ``x^^^``.

There is a single postfix operator:

``x^``
  Equivalent to ``x^1``, the first parent of each changeset in x.


The following predicates are supported:

.. predicatesmarker

New predicates (known as "aliases") can be defined, using any combination of
existing predicates or other aliases. An alias definition looks like::

  <alias> = <definition>

in the ``revsetalias`` section of a Mercurial configuration file. Arguments
of the form `a1`, `a2`, etc. are substituted from the alias into the
definition.

For example,

::

  [revsetalias]
  h = heads()
  d(s) = sort(s, date)
  rs(s, k) = reverse(sort(s, k))

defines three aliases, ``h``, ``d``, and ``rs``. ``rs(0:tip, author)`` is
exactly equivalent to ``reverse(sort(0:tip, author))``.

An infix operator ``##`` can concatenate strings and identifiers into
one string. For example::

  [revsetalias]
  issue(a1) = grep(r'\bissue[ :]?' ## a1 ## r'\b|\bbug\(' ## a1 ## r'\)')

``issue(1234)`` is equivalent to ``grep(r'\bissue[ :]?1234\b|\bbug\(1234\)')``
in this case. This matches against all of "issue 1234", "issue:1234",
"issue1234" and "bug(1234)".

All other prefix, infix and postfix operators have lower priority than
``##``. For example, ``a1 ## a2~2`` is equivalent to ``(a1 ## a2)~2``.

Command line equivalents for :hg:`log`::

  -f    ->  ::.
  -d x  ->  date(x)
  -k x  ->  keyword(x)
  -m    ->  merge()
  -u x  ->  user(x)
  -b x  ->  branch(x)
  -P x  ->  !::x
  -l x  ->  limit(expr, x)

Some sample queries:

- Changesets on the default branch::

    hg log -r "branch(default)"

- Changesets on the default branch since tag 1.5 (excluding merges)::

    hg log -r "branch(default) and 1.5:: and not merge()"

- Open branch heads::

    hg log -r "head() and not closed()"

- Changesets between tags 1.3 and 1.5 mentioning "bug" that affect
  ``hgext/*``::

    hg log -r "1.3::1.5 and keyword(bug) and file('hgext/*')"

- Changesets committed in May 2008, sorted by user::

    hg log -r "sort(date('May 2008'), user)"

- Changesets mentioning "bug" or "issue" that are not in a tagged
  release::

    hg log -r "(keyword(bug) or keyword(issue)) and not ancestors(tag())"