revsetlang: build optimized tree by helper function

This should make optimize() more readable, but it doubles the parsing cost.

  (original)
  $ python -m timeit -n10000 -s 'from mercurial import revsetlang as L' \
  'L.optimize(L.analyze(L.parse("::tip")))'
  10000 loops, best of 3: 18.1 usec per loop

  (this patch)
  $ python -m timeit -n10000 -s 'from mercurial import revsetlang as L' \
  'L._treecache.clear(); L.optimize(L.analyze(L.parse("::tip")))'
  10000 loops, best of 3: 48.4 usec per loop

30usec isn't dominant compared to the revset evaluation, but that is a cost.
That's why a parsed tree is cached, which can benefit in hgweb or chg server.

mercurial/revsetlang.py

@@ -1,671 +1,688 b''
 # revsetlang.py - parser, tokenizer and utility for revision set language
 #
 # Copyright 2010 Matt Mackall <[e-mail]>
 #
 # This software may be used and distributed according to the terms of the
 # GNU General Public License version 2 or any later version.
 
 from __future__ import absolute_import
 
 import string
 
 from .i18n import _
 from . import (
     error,
     node,
     parser,
     pycompat,
     util,
 )
 
 elements = {
     # token-type: binding-strength, primary, prefix, infix, suffix
     "(": (21, None, ("group", 1, ")"), ("func", 1, ")"), None),
     "[": (21, None, None, ("subscript", 1, "]"), None),
     "#": (21, None, None, ("relation", 21), None),
     "##": (20, None, None, ("_concat", 20), None),
     "~": (18, None, None, ("ancestor", 18), None),
     "^": (18, None, None, ("parent", 18), "parentpost"),
     "-": (5, None, ("negate", 19), ("minus", 5), None),
     "::": (17, None, ("dagrangepre", 17), ("dagrange", 17), "dagrangepost"),
     "..": (17, None, ("dagrangepre", 17), ("dagrange", 17), "dagrangepost"),
     ":": (15, "rangeall", ("rangepre", 15), ("range", 15), "rangepost"),
     "not": (10, None, ("not", 10), None, None),
     "!": (10, None, ("not", 10), None, None),
     "and": (5, None, None, ("and", 5), None),
     "&": (5, None, None, ("and", 5), None),
     "%": (5, None, None, ("only", 5), "onlypost"),
     "or": (4, None, None, ("or", 4), None),
     "|": (4, None, None, ("or", 4), None),
     "+": (4, None, None, ("or", 4), None),
     "=": (3, None, None, ("keyvalue", 3), None),
     ",": (2, None, None, ("list", 2), None),
     ")": (0, None, None, None, None),
     "]": (0, None, None, None, None),
     "symbol": (0, "symbol", None, None, None),
     "string": (0, "string", None, None, None),
     "end": (0, None, None, None, None),
 }
 
 keywords = {'and', 'or', 'not'}
 
 _quoteletters = {'"', "'"}
 _simpleopletters = set(pycompat.iterbytestr("()[]#:=,-|&+!~^%"))
 
 # default set of valid characters for the initial letter of symbols
 _syminitletters = set(pycompat.iterbytestr(
     string.ascii_letters.encode('ascii') +
     string.digits.encode('ascii') +
     '._@')) | set(map(pycompat.bytechr, xrange(128, 256)))
 
 # default set of valid characters for non-initial letters of symbols
 _symletters = _syminitletters | set(pycompat.iterbytestr('-/'))
 
 def tokenize(program, lookup=None, syminitletters=None, symletters=None):
     '''
     Parse a revset statement into a stream of tokens
 
     ``syminitletters`` is the set of valid characters for the initial
     letter of symbols.
 
     By default, character ``c`` is recognized as valid for initial
     letter of symbols, if ``c.isalnum() or c in '._@' or ord(c) > 127``.
 
     ``symletters`` is the set of valid characters for non-initial
     letters of symbols.
 
     By default, character ``c`` is recognized as valid for non-initial
     letters of symbols, if ``c.isalnum() or c in '-._/@' or ord(c) > 127``.
 
     Check that @ is a valid unquoted token character (issue3686):
     >>> list(tokenize("@::"))
     [('symbol', '@', 0), ('::', None, 1), ('end', None, 3)]
 
     '''
     program = pycompat.bytestr(program)
     if syminitletters is None:
         syminitletters = _syminitletters
     if symletters is None:
         symletters = _symletters
 
     if program and lookup:
         # attempt to parse old-style ranges first to deal with
         # things like old-tag which contain query metacharacters
         parts = program.split(':', 1)
         if all(lookup(sym) for sym in parts if sym):
             if parts[0]:
                 yield ('symbol', parts[0], 0)
             if len(parts) > 1:
                 s = len(parts[0])
                 yield (':', None, s)
                 if parts[1]:
                     yield ('symbol', parts[1], s + 1)
             yield ('end', None, len(program))
             return
 
     pos, l = 0, len(program)
     while pos < l:
         c = program[pos]
         if c.isspace(): # skip inter-token whitespace
             pass
         elif c == ':' and program[pos:pos + 2] == '::': # look ahead carefully
             yield ('::', None, pos)
             pos += 1 # skip ahead
         elif c == '.' and program[pos:pos + 2] == '..': # look ahead carefully
             yield ('..', None, pos)
             pos += 1 # skip ahead
         elif c == '#' and program[pos:pos + 2] == '##': # look ahead carefully
             yield ('##', None, pos)
             pos += 1 # skip ahead
         elif c in _simpleopletters: # handle simple operators
             yield (c, None, pos)
         elif (c in _quoteletters or c == 'r' and
               program[pos:pos + 2] in ("r'", 'r"')): # handle quoted strings
             if c == 'r':
                 pos += 1
                 c = program[pos]
                 decode = lambda x: x
             else:
                 decode = parser.unescapestr
             pos += 1
             s = pos
             while pos < l: # find closing quote
                 d = program[pos]
                 if d == '\\': # skip over escaped characters
                     pos += 2
                     continue
                 if d == c:
                     yield ('string', decode(program[s:pos]), s)
                     break
                 pos += 1
             else:
                 raise error.ParseError(_("unterminated string"), s)
         # gather up a symbol/keyword
         elif c in syminitletters:
             s = pos
             pos += 1
             while pos < l: # find end of symbol
                 d = program[pos]
                 if d not in symletters:
                     break
                 if d == '.' and program[pos - 1] == '.': # special case for ..
                     pos -= 1
                     break
                 pos += 1
             sym = program[s:pos]
             if sym in keywords: # operator keywords
                 yield (sym, None, s)
             elif '-' in sym:
                 # some jerk gave us foo-bar-baz, try to check if it's a symbol
                 if lookup and lookup(sym):
                     # looks like a real symbol
                     yield ('symbol', sym, s)
                 else:
                     # looks like an expression
                     parts = sym.split('-')
                     for p in parts[:-1]:
                         if p: # possible consecutive -
                             yield ('symbol', p, s)
                         s += len(p)
                         yield ('-', None, pos)
                         s += 1
                     if parts[-1]: # possible trailing -
                         yield ('symbol', parts[-1], s)
             else:
                 yield ('symbol', sym, s)
             pos -= 1
         else:
             raise error.ParseError(_("syntax error in revset '%s'") %
                                    program, pos)
         pos += 1
     yield ('end', None, pos)
 
 # helpers
 
 _notset = object()
 
 def getsymbol(x):
     if x and x[0] == 'symbol':
         return x[1]
     raise error.ParseError(_('not a symbol'))
 
 def getstring(x, err):
     if x and (x[0] == 'string' or x[0] == 'symbol'):
         return x[1]
     raise error.ParseError(err)
 
 def getinteger(x, err, default=_notset):
     if not x and default is not _notset:
         return default
     try:
         return int(getstring(x, err))
     except ValueError:
         raise error.ParseError(err)
 
 def getboolean(x, err):
     value = util.parsebool(getsymbol(x))
     if value is not None:
         return value
     raise error.ParseError(err)
 
 def getlist(x):
     if not x:
         return []
     if x[0] == 'list':
         return list(x[1:])
     return [x]
 
 def getrange(x, err):
     if not x:
         raise error.ParseError(err)
     op = x[0]
     if op == 'range':
         return x[1], x[2]
     elif op == 'rangepre':
         return None, x[1]
     elif op == 'rangepost':
         return x[1], None
     elif op == 'rangeall':
         return None, None
     raise error.ParseError(err)
 
 def getargs(x, min, max, err):
     l = getlist(x)
     if len(l) < min or (max >= 0 and len(l) > max):
         raise error.ParseError(err)
     return l
 
 def getargsdict(x, funcname, keys):
     return parser.buildargsdict(getlist(x), funcname, parser.splitargspec(keys),
                                 keyvaluenode='keyvalue', keynode='symbol')
 
+# cache of {spec: raw parsed tree} built internally
+_treecache = {}
+
+def _cachedtree(spec):
+    # thread safe because parse() is reentrant and dict.__setitem__() is atomic
+    tree = _treecache.get(spec)
+    if tree is None:
+        _treecache[spec] = tree = parse(spec)
+    return tree
+
+def _build(tmplspec, *repls):
+    """Create raw parsed tree from a template revset statement
+
+    >>> _build('f(_) and _', ('string', '1'), ('symbol', '2'))
+    ('and', ('func', ('symbol', 'f'), ('string', '1')), ('symbol', '2'))
+    """
+    template = _cachedtree(tmplspec)
+    return parser.buildtree(template, ('symbol', '_'), *repls)
+
 def _isnamedfunc(x, funcname):
     """Check if given tree matches named function"""
     return x and x[0] == 'func' and getsymbol(x[1]) == funcname
 
 def _isposargs(x, n):
     """Check if given tree is n-length list of positional arguments"""
     l = getlist(x)
     return len(l) == n and all(y and y[0] != 'keyvalue' for y in l)
 
 def _matchnamedfunc(x, funcname):
     """Return args tree if given tree matches named function; otherwise None
 
     This can't be used for testing a nullary function since its args tree
     is also None. Use _isnamedfunc() instead.
     """
     if not _isnamedfunc(x, funcname):
         return
     return x[2]
 
 def _matchonly(revs, bases):
     """
     >>> f = lambda *args: _matchonly(*map(parse, args))
     >>> f('ancestors(A)', 'not ancestors(B)')
     ('list', ('symbol', 'A'), ('symbol', 'B'))
     """
     ta = _matchnamedfunc(revs, 'ancestors')
     tb = bases and bases[0] == 'not' and _matchnamedfunc(bases[1], 'ancestors')
     if _isposargs(ta, 1) and _isposargs(tb, 1):
         return ('list', ta, tb)
 
 def _fixops(x):
     """Rewrite raw parsed tree to resolve ambiguous syntax which cannot be
     handled well by our simple top-down parser"""
     if not isinstance(x, tuple):
         return x
 
     op = x[0]
     if op == 'parent':
         # x^:y means (x^) : y, not x ^ (:y)
         # x^:  means (x^) :,   not x ^ (:)
         post = ('parentpost', x[1])
         if x[2][0] == 'dagrangepre':
             return _fixops(('dagrange', post, x[2][1]))
         elif x[2][0] == 'rangepre':
             return _fixops(('range', post, x[2][1]))
         elif x[2][0] == 'rangeall':
             return _fixops(('rangepost', post))
     elif op == 'or':
         # make number of arguments deterministic:
         # x + y + z -> (or x y z) -> (or (list x y z))
         return (op, _fixops(('list',) + x[1:]))
     elif op == 'subscript' and x[1][0] == 'relation':
         # x#y[z] ternary
         return _fixops(('relsubscript', x[1][1], x[1][2], x[2]))
 
     return (op,) + tuple(_fixops(y) for y in x[1:])
 
 def _analyze(x):
     if x is None:
         return x
 
     op = x[0]
     if op == 'minus':
-        return _analyze(('and', x[1], ('not', x[2])))
+        return _analyze(_build('_ and not _', *x[1:]))
     elif op == 'only':
-        t = ('func', ('symbol', 'only'), ('list', x[1], x[2]))
-        return _analyze(t)
+        return _analyze(_build('only(_, _)', *x[1:]))
     elif op == 'onlypost':
-        return _analyze(('func', ('symbol', 'only'), x[1]))
+        return _analyze(_build('only(_)', x[1]))
     elif op == 'dagrangepre':
-        return _analyze(('func', ('symbol', 'ancestors'), x[1]))
+        return _analyze(_build('ancestors(_)', x[1]))
     elif op == 'dagrangepost':
-        return _analyze(('func', ('symbol', 'descendants'), x[1]))
+        return _analyze(_build('descendants(_)', x[1]))
     elif op == 'negate':
         s = getstring(x[1], _("can't negate that"))
         return _analyze(('string', '-' + s))
     elif op in ('string', 'symbol'):
         return x
     elif op == 'rangeall':
         return (op, None)
     elif op in {'or', 'not', 'rangepre', 'rangepost', 'parentpost'}:
         return (op, _analyze(x[1]))
     elif op == 'group':
         return _analyze(x[1])
     elif op in {'and', 'dagrange', 'range', 'parent', 'ancestor', 'relation',
                 'subscript'}:
         ta = _analyze(x[1])
         tb = _analyze(x[2])
         return (op, ta, tb)
     elif op == 'relsubscript':
         ta = _analyze(x[1])
         tb = _analyze(x[2])
         tc = _analyze(x[3])
         return (op, ta, tb, tc)
     elif op == 'list':
         return (op,) + tuple(_analyze(y) for y in x[1:])
     elif op == 'keyvalue':
         return (op, x[1], _analyze(x[2]))
     elif op == 'func':
         return (op, x[1], _analyze(x[2]))
     raise ValueError('invalid operator %r' % op)
 
 def analyze(x):
     """Transform raw parsed tree to evaluatable tree which can be fed to
     optimize() or getset()
 
     All pseudo operations should be mapped to real operations or functions
     defined in methods or symbols table respectively.
     """
     return _analyze(x)
 
 def _optimize(x, small):
     if x is None:
         return 0, x
 
     smallbonus = 1
     if small:
         smallbonus = .5
 
     op = x[0]
     if op in ('string', 'symbol'):
         return smallbonus, x # single revisions are small
     elif op == 'and':
         wa, ta = _optimize(x[1], True)
         wb, tb = _optimize(x[2], True)
         w = min(wa, wb)
 
         # (::x and not ::y)/(not ::y and ::x) have a fast path
-        tm = _matchonly(ta, tb) or _matchonly(tb, ta)
-        if tm:
-            return w, ('func', ('symbol', 'only'), tm)
+        m = _matchonly(ta, tb) or _matchonly(tb, ta)
+        if m:
+            return w, _build('only(_, _)', *m[1:])
 
         if tb is not None and tb[0] == 'not':
             return wa, ('difference', ta, tb[1])
         if wa > wb:
             op = 'andsmally'
         return w, (op, ta, tb)
     elif op == 'or':
         # fast path for machine-generated expression, that is likely to have
         # lots of trivial revisions: 'a + b + c()' to '_list(a b) + c()'
         ws, ts, ss = [], [], []
         def flushss():
             if not ss:
                 return
             if len(ss) == 1:
                 w, t = ss[0]
             else:
                 s = '\0'.join(t[1] for w, t in ss)
-                y = ('func', ('symbol', '_list'), ('string', s))
+                y = _build('_list(_)', ('string', s))
                 w, t = _optimize(y, False)
             ws.append(w)
             ts.append(t)
             del ss[:]
         for y in getlist(x[1]):
             w, t = _optimize(y, False)
             if t is not None and (t[0] == 'string' or t[0] == 'symbol'):
                 ss.append((w, t))
                 continue
             flushss()
             ws.append(w)
             ts.append(t)
         flushss()
         if len(ts) == 1:
             return ws[0], ts[0] # 'or' operation is fully optimized out
         return max(ws), (op, ('list',) + tuple(ts))
     elif op == 'not':
         # Optimize not public() to _notpublic() because we have a fast version
         if x[1][:3] == ('func', ('symbol', 'public'), None):
-            newsym = ('func', ('symbol', '_notpublic'), None)
-            o = _optimize(newsym, not small)
+            o = _optimize(_build('_notpublic()'), not small)
             return o[0], o[1]
         else:
             o = _optimize(x[1], not small)
             return o[0], (op, o[1])
     elif op == 'rangeall':
         return smallbonus, x
     elif op in ('rangepre', 'rangepost', 'parentpost'):
         o = _optimize(x[1], small)
         return o[0], (op, o[1])
     elif op in ('dagrange', 'range'):
         wa, ta = _optimize(x[1], small)
         wb, tb = _optimize(x[2], small)
         return wa + wb, (op, ta, tb)
     elif op in ('parent', 'ancestor', 'relation', 'subscript'):
         w, t = _optimize(x[1], small)
         return w, (op, t, x[2])
     elif op == 'relsubscript':
         w, t = _optimize(x[1], small)
         return w, (op, t, x[2], x[3])
     elif op == 'list':
         ws, ts = zip(*(_optimize(y, small) for y in x[1:]))
         return sum(ws), (op,) + ts
     elif op == 'keyvalue':
         w, t = _optimize(x[2], small)
         return w, (op, x[1], t)
     elif op == 'func':
         f = getsymbol(x[1])
         wa, ta = _optimize(x[2], small)
         if f in ('author', 'branch', 'closed', 'date', 'desc', 'file', 'grep',
                  'keyword', 'outgoing', 'user', 'destination'):
             w = 10 # slow
         elif f in ('modifies', 'adds', 'removes'):
             w = 30 # slower
         elif f == "contains":
             w = 100 # very slow
         elif f == "ancestor":
             w = 1 * smallbonus
         elif f in ('reverse', 'limit', 'first', 'wdir', '_intlist'):
             w = 0
         elif f == "sort":
             w = 10 # assume most sorts look at changelog
         else:
             w = 1
         return w + wa, (op, x[1], ta)
     raise ValueError('invalid operator %r' % op)
 
 def optimize(tree):
     """Optimize evaluatable tree
 
     All pseudo operations should be transformed beforehand.
     """
     _weight, newtree = _optimize(tree, small=True)
     return newtree
 
 # the set of valid characters for the initial letter of symbols in
 # alias declarations and definitions
 _aliassyminitletters = _syminitletters | set(pycompat.sysstr('$'))
 
 def _parsewith(spec, lookup=None, syminitletters=None):
     """Generate a parse tree of given spec with given tokenizing options
 
     >>> _parsewith('foo($1)', syminitletters=_aliassyminitletters)
     ('func', ('symbol', 'foo'), ('symbol', '$1'))
     >>> _parsewith('$1')
     Traceback (most recent call last):
       ...
     ParseError: ("syntax error in revset '$1'", 0)
     >>> _parsewith('foo bar')
     Traceback (most recent call last):
       ...
     ParseError: ('invalid token', 4)
     """
     p = parser.parser(elements)
     tree, pos = p.parse(tokenize(spec, lookup=lookup,
                                  syminitletters=syminitletters))
     if pos != len(spec):
         raise error.ParseError(_('invalid token'), pos)
     return _fixops(parser.simplifyinfixops(tree, ('list', 'or')))
 
 class _aliasrules(parser.basealiasrules):
     """Parsing and expansion rule set of revset aliases"""
     _section = _('revset alias')
 
     @staticmethod
     def _parse(spec):
         """Parse alias declaration/definition ``spec``
 
         This allows symbol names to use also ``$`` as an initial letter
         (for backward compatibility), and callers of this function should
         examine whether ``$`` is used also for unexpected symbols or not.
         """
         return _parsewith(spec, syminitletters=_aliassyminitletters)
 
     @staticmethod
     def _trygetfunc(tree):
         if tree[0] == 'func' and tree[1][0] == 'symbol':
             return tree[1][1], getlist(tree[2])
 
 def expandaliases(tree, aliases, warn=None):
     """Expand aliases in a tree, aliases is a list of (name, value) tuples"""
     aliases = _aliasrules.buildmap(aliases)
     tree = _aliasrules.expand(aliases, tree)
     # warn about problematic (but not referred) aliases
     if warn is not None:
         for name, alias in sorted(aliases.iteritems()):
             if alias.error and not alias.warned:
                 warn(_('warning: %s\n') % (alias.error))
                 alias.warned = True
     return tree
 
 def foldconcat(tree):
     """Fold elements to be concatenated by `##`
     """
     if not isinstance(tree, tuple) or tree[0] in ('string', 'symbol'):
         return tree
     if tree[0] == '_concat':
         pending = [tree]
         l = []
         while pending:
             e = pending.pop()
             if e[0] == '_concat':
                 pending.extend(reversed(e[1:]))
             elif e[0] in ('string', 'symbol'):
                 l.append(e[1])
             else:
                 msg = _("\"##\" can't concatenate \"%s\" element") % (e[0])
                 raise error.ParseError(msg)
         return ('string', ''.join(l))
     else:
         return tuple(foldconcat(t) for t in tree)
 
 def parse(spec, lookup=None):
     return _parsewith(spec, lookup=lookup)
 
 def _quote(s):
     r"""Quote a value in order to make it safe for the revset engine.
 
     >>> _quote('asdf')
     "'asdf'"
     >>> _quote("asdf'\"")
     '\'asdf\\\'"\''
     >>> _quote('asdf\'')
     "'asdf\\''"
     >>> _quote(1)
     "'1'"
     """
     return "'%s'" % util.escapestr(pycompat.bytestr(s))
 
 def formatspec(expr, *args):
     '''
     This is a convenience function for using revsets internally, and
     escapes arguments appropriately. Aliases are intentionally ignored
     so that intended expression behavior isn't accidentally subverted.
 
     Supported arguments:
 
     %r = revset expression, parenthesized
     %d = int(arg), no quoting
     %s = string(arg), escaped and single-quoted
     %b = arg.branch(), escaped and single-quoted
     %n = hex(arg), single-quoted
     %% = a literal '%'
 
     Prefixing the type with 'l' specifies a parenthesized list of that type.
 
     >>> formatspec('%r:: and %lr', '10 or 11', ("this()", "that()"))
     '(10 or 11):: and ((this()) or (that()))'
     >>> formatspec('%d:: and not %d::', 10, 20)
     '10:: and not 20::'
     >>> formatspec('%ld or %ld', [], [1])
     "_list('') or 1"
     >>> formatspec('keyword(%s)', 'foo\\xe9')
     "keyword('foo\\\\xe9')"
     >>> b = lambda: 'default'
     >>> b.branch = b
     >>> formatspec('branch(%b)', b)
     "branch('default')"
     >>> formatspec('root(%ls)', ['a', 'b', 'c', 'd'])
     "root(_list('a\\x00b\\x00c\\x00d'))"
     '''
 
     def argtype(c, arg):
         if c == 'd':
             return '%d' % int(arg)
         elif c == 's':
             return _quote(arg)
         elif c == 'r':
             parse(arg) # make sure syntax errors are confined
             return '(%s)' % arg
         elif c == 'n':
             return _quote(node.hex(arg))
         elif c == 'b':
             return _quote(arg.branch())
 
     def listexp(s, t):
         l = len(s)
         if l == 0:
             return "_list('')"
         elif l == 1:
             return argtype(t, s[0])
         elif t == 'd':
             return "_intlist('%s')" % "\0".join('%d' % int(a) for a in s)
         elif t == 's':
             return "_list('%s')" % "\0".join(s)
         elif t == 'n':
             return "_hexlist('%s')" % "\0".join(node.hex(a) for a in s)
         elif t == 'b':
             return "_list('%s')" % "\0".join(a.branch() for a in s)
 
         m = l // 2
         return '(%s or %s)' % (listexp(s[:m], t), listexp(s[m:], t))
 
     expr = pycompat.bytestr(expr)
     ret = ''
     pos = 0
     arg = 0
     while pos < len(expr):
         c = expr[pos]
         if c == '%':
             pos += 1
             d = expr[pos]
             if d == '%':
                 ret += d
             elif d in 'dsnbr':
                 ret += argtype(d, args[arg])
                 arg += 1
             elif d == 'l':
                 # a list of some type
                 pos += 1
                 d = expr[pos]
                 ret += listexp(list(args[arg]), d)
                 arg += 1
             else:
                 raise error.Abort(_('unexpected revspec format character %s')
                                   % d)
         else:
             ret += c
         pos += 1
 
     return ret
 
 def prettyformat(tree):
     return parser.prettyformat(tree, ('string', 'symbol'))
 
 def depth(tree):
     if isinstance(tree, tuple):
         return max(map(depth, tree)) + 1
     else:
         return 0
 
 def funcsused(tree):
     if not isinstance(tree, tuple) or tree[0] in ('string', 'symbol'):
         return set()
     else:
         funcs = set()
         for s in tree[1:]:
             funcs |= funcsused(s)
         if tree[0] == 'func':
             funcs.add(tree[1][1])
         return funcs