Add a Python parser for GN types to gn_helpers.

build/gn_helpers.py

Index: build/gn_helpers.py
diff --git a/build/gn_helpers.py b/build/gn_helpers.py
index 3b0647d9a5b7ea9ab60fa2c02fab63c3f39c85d2..592a0f790a651964cfff2365cdf3ad0d4f329b17 100644
--- a/build/gn_helpers.py
+++ b/build/gn_helpers.py
@@ -2,8 +2,10 @@
 # Use of this source code is governed by a BSD-style license that can be
 # found in the LICENSE file.
 
-"""Helper functions useful when writing scripts that are run from GN's
-exec_script function."""
+"""Helper functions useful when writing scripts that integrate with GN.
+
+The main functions are ToGNString and FromGNString which convert between
+serialized GN veriables and Python variables."""
 
 class GNException(Exception):
   pass
@@ -18,7 +20,14 @@ def ToGNString(value, allow_dicts = True):
   if isinstance(value, str):
     if value.find('\n') >= 0:
       raise GNException("Trying to print a string with a newline in it.")
-    return '"' + value.replace('"', '\\"') + '"'
+    return '"' + \
+        value.replace('\\', '\\\\').replace('"', '\\"').replace('$', '\\$') + \
+        '"'
+
+  if isinstance(value, bool):
+    if value:
+      return "true"
+    return "false"
 
   if isinstance(value, list):
     return '[ %s ]' % ', '.join(ToGNString(v) for v in value)
@@ -37,3 +46,227 @@ def ToGNString(value, allow_dicts = True):
     return str(value)
 
   raise GNException("Unsupported type when printing to GN.")
+
+
+def FromGNString(input):
+  """Converts the input string from a GN serialized value to Python values.
+
+  For details on supported types see GNValueParser.Parse() below.
+
+  If your GN script did:
+    something = [ "file1", "file2" ]
+    args = [ "--values=$something" ]
+  The command line would look something like:
+    --values="[ \"file1\", \"file2\" ]"
+  Which when interpreted as a command line gives the value:
+    [ "file1", "file2" ]
+
+  You can parse this into a Python list using GN rules with:
+    input_values = FromGNValues(options.values)
+  Although the Python 'ast' module will parse many forms of such input, it
+  will not handle GN escaping properly, nor GN booleans. You should use this
+  function instead.
+
+
+  A NOTE ON STRING HANDLING:
+
+  If you just pass a string on the command line to your Python script, or use
+  string interpolation on a string variable, the strings will not be quoted:
+    str = "asdf"
+    args = [ str, "--value=$str" ]
+  Will yield the command line:
+    asdf --value=asdf
+  The unquoted asdf string will not be valid input to this function, which
+  accepts only quoted strings like GN scripts. In such cases, you can just use
+  the Python string literal directly.
+
+  The main use cases for this is for other types, in particular lists. When
+  using string interpolation on a list (as in the top example) the embedded
+  strings will be quoted and escaped according to GN rules so the list can be
+  re-parsed to get the same result."""
+  parser = GNValueParser(input)
+  return parser.Parse()
+
+
+def UnescapeGNString(value):

[Dirk Pranke, 2016/01/22 00:34:02]

Is this supposed to be a public function, or should it be private to this file?

[brettw, 2016/01/22 18:12:14]

I was thinking public since one could conceive of needing such a helper function
in the future.

+  """Given a string with GN escaping, returns the unescaped string.
+
+  Be careful not to feed with input from a Python parsing function like
+  'ast' because it will do Python unescaping, which will be incorrect when
+  fed into the GN unescaper."""
+  result = ''
+  i = 0
+  while i < len(value):
+    if value[i] == '\\':
+      if i < len(value) - 1:
+        next_char = value[i + 1]
+        if next_char == '$' or next_char == '"' or next_char == '\\':

[Dirk Pranke, 2016/01/22 00:34:02]

nit: maybe

  if next_char in ('$', '"', '\\'):

[brettw, 2016/01/22 18:12:15]

Done.

+          # These are the escaped characters GN supports.
+          result += next_char
+          i += 1
+        else:
+          # Any other backslash is a litera.

[Dirk Pranke, 2016/01/22 00:34:02]

typo: s/litera/literal/ .

+          result += '\\'
+    else:
+      result += value[i]
+    i += 1
+  return result
+
+
+class GNValueParser:

[Dirk Pranke, 2016/01/22 00:34:02]

nit: inherit from object, like: 

class GNValueParser(object):

Also, is this object supposed to be public?

[brettw, 2016/01/22 18:12:15]

Yes, if you want to parse a specific type or something.

+  """Duplicates GN parsing of values and converts to Python types.
+
+  Normally you would use the wrapper function FromGNValue() below.
+
+  If you expect input as a specific type, you can also call one of the Parse*
+  functions directly. All functions throw GNException on invalid input. """
+  def __init__(self, string):
+    self.input = string
+    self.cur = 0
+
+  def IsDone(self):
+    return self.cur == len(self.input)
+
+  def ConsumeWhitespace(self):
+    while not self.IsDone() and self.input[self.cur] in ' \t\n':
+      self.cur += 1
+
+  def Parse(self):
+    """Converts a string representing a printed GN value to the Python type.
+
+    See additional usage notes on FromGNString above.
+
+    - GN booleans ('true', 'false') will be converted to Python booleans.
+
+    - GN numbers ('123') will be converted to Python numbers.
+
+    - GN strings (double-quoted as in '"asdf"') will be converted to Python
+      strings with GN escaping rules. GN string interpolation (embedded
+      variables preceeded by $) are not supported and will be returned as
+      literals.
+
+    - GN lists ('[1, "asdf", 3]') will be converted to Python lists.
+
+    - GN scopes ('{ ... }') are not supported."""
+    result = self.ParseAllowTrailing()
+    self.ConsumeWhitespace()
+    if not self.IsDone():
+      raise GNException("Trailing input after parsing:\n  " +
+                        self.input[self.cur:])
+    return result
+
+  def ParseAllowTrailing(self):

[Dirk Pranke, 2016/01/22 00:34:02]

if this and the following functions are not public, they should be named
_ParseAllowTrailing() et al, and some of your assertions are really kinda
unneeded at that point.

+    """Internal version of Parse that doesn't check for trailing stuff."""
+    self.ConsumeWhitespace()
+    if self.IsDone():
+      raise GNException("Expected input to parse.")
+
+    next_char = self.input[self.cur]
+    if next_char == '[':
+      return self.ParseList()
+    elif next_char in "-0123456789":

[Dirk Pranke, 2016/01/22 00:34:02]

nit: consider making this a helper function: IsDigitOrMinus() or some such
thing, and re-use here and on line 186?

You might also rewrite this as next_char.isdigit() or next_char == '-', but I
admit that I'm not sure if that's either faster or clearer.

+      return self.ParseNumber()
+    elif next_char == '"':
+      return self.ParseString()
+    elif self.ConstantFollows('true'):
+      return True
+    elif self.ConstantFollows('false'):
+      return False
+    else:
+      raise GNException("Unexpected token: " + self.input[self.cur:])
+
+  def ParseNumber(self):
+    self.ConsumeWhitespace()
+    if self.IsDone():
+      raise GNException('Expected number but got nothing.')
+
+    begin = self.cur
+
+    # The first character can include a negative sign.
+    if not self.IsDone() and self.input[self.cur] in '-0123456789':
+      self.cur += 1
+    while not self.IsDone() and self.input[self.cur] in '0123456789':

[Dirk Pranke, 2016/01/22 00:34:02]

maybe self.input[self.cur].isdigit() ? maybe not ...

alternatively, why not match against a regex? to better mirror the c++ code,
maybe?

[brettw, 2016/01/22 18:12:14]

I don't think in regex's and the code is longer. I did isdigit()

+      self.cur += 1
+
+    number_string = self.input[begin:self.cur]
+    if not len(number_string) or number_string == '-':
+      raise GNException("Not a valid number.")
+    return int(number_string)
+
+  def ParseString(self):
+    self.ConsumeWhitespace()
+    if self.IsDone():
+      raise GNException('Expected string but got nothing.')
+
+    if self.input[self.cur] != '"':
+      raise GNException('Expected string beginning in a " but got:\n  ' +
+                        self.input[self.cur:])
+    self.cur += 1  # Skip over quote.
+
+    begin = self.cur
+    while not self.IsDone() and self.input[self.cur] != '"':
+      if self.input[self.cur] == '\\':
+        self.cur += 1  # Skip over the backslash.
+        if self.IsDone():
+          raise GNException("String ends in a backslash in:\n  " +
+                            self.input)
+      self.cur += 1
+
+    if self.IsDone():
+      raise GNException('Unterminated string:\n  ' + self.input[begin:])
+
+    end = self.cur
+    self.cur += 1  # Consume trailing ".
+
+    return UnescapeGNString(self.input[begin:end])
+
+  def ParseList(self):
+    self.ConsumeWhitespace()
+    if self.IsDone():
+      raise GNException('Expected string but got nothing.')

[Dirk Pranke, 2016/01/22 00:34:02]

'Expected list' instead?

[brettw, 2016/01/22 18:12:15]

Done.

+
+    # Skip over opening '['.
+    if self.input[self.cur] != '[':
+      raise GNException("Expected [ for list but got:\n  " +
+                        self.input[self.cur:])
+    self.cur += 1
+    self.ConsumeWhitespace()
+    if self.IsDone():
+      raise GNException("Unterminated list:\n  " + self.input)
+
+    list_result = []
+    previous_had_trailing_comma = True
+    while not self.IsDone():
+      if self.input[self.cur] == ']':
+        self.cur += 1  # Skip over ']'.
+        return list_result
+
+      if not previous_had_trailing_comma:
+        raise GNException("List items not separated by comma.")
+
+      list_result += [ self.ParseAllowTrailing() ]
+      self.ConsumeWhitespace()
+      if self.IsDone():
+        break
+
+      # Consume comma if there is one.
+      previous_had_trailing_comma = self.input[self.cur] == ','
+      if previous_had_trailing_comma:
+        # Consume comma.
+        self.cur += 1
+        self.ConsumeWhitespace()
+
+    raise GNException("Unterminated list:\n  " + self.input)
+
+  def ConstantFollows(self, constant):

[Dirk Pranke, 2016/01/22 00:34:02]

nit: _ConstantFollows() if this isn't a public function.

[brettw, 2016/01/22 18:12:14]

Done.

+    """Returns true if the given constant follows immediately at the current
+    location in the input. If it does, the text is consumed and the function
+    returns true. Otherwise, returns false and the current position is
+    unchanged."""
+    end = self.cur + len(constant)
+    if end >= len(self.input):
+      return False  # Not enough room.
+    if self.input[self.cur:end] == constant:
+      self.cur = end
+      return True
+    return False