summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorDieter Verfaillie <dieterv@optionexplicit.be>2013-06-21 17:40:29 +0200
committerDieter Verfaillie <dieterv@optionexplicit.be>2013-10-08 20:54:34 +0200
commitc9c7de5d359bf661fc57639a7b53b07045a4e092 (patch)
treeed54a5da3286ee40c71376ec8b0d805b813bdc92
parentd8998d68d3dd03e7e8bfc87ec4464512ed705ac2 (diff)
downloadgobject-introspection-c9c7de5d359bf661fc57639a7b53b07045a4e092.tar.gz
giscanner: rename "comment" to "description"
Simply because they are called description fields on GTK-Doc parts...
-rw-r--r--giscanner/annotationparser.py42
-rw-r--r--giscanner/maintransformer.py10
-rw-r--r--tests/scanner/annotationparser/test_parser.py12
3 files changed, 32 insertions, 32 deletions
diff --git a/giscanner/annotationparser.py b/giscanner/annotationparser.py
index 4faa2d38..f6f74000 100644
--- a/giscanner/annotationparser.py
+++ b/giscanner/annotationparser.py
@@ -458,7 +458,7 @@ class DocBlock(object):
self.annotations = DocAnnotations()
self.value = None
self.tags = OrderedDict()
- self.comment = None
+ self.description = None
self.params = OrderedDict()
self.position = None
@@ -482,9 +482,9 @@ class DocBlock(object):
lines[0] += annotations
for param in self.params.values():
lines.append(param.to_gtk_doc_param())
- if self.comment:
+ if self.description:
lines.append('')
- for l in self.comment.split('\n'):
+ for l in self.description.split('\n'):
lines.append(l)
if self.tags:
lines.append('')
@@ -516,7 +516,7 @@ class DocTag(object):
self.block = block
self.name = name
self.annotations = DocAnnotations()
- self.comment = None
+ self.description = None
self.value = ''
self.position = None
@@ -625,12 +625,12 @@ class DocTag(object):
return self.value
def to_gtk_doc_param(self):
- return '@%s: %s%s' % (self.name, self._get_gtk_doc_value(), self.comment)
+ return '@%s: %s%s' % (self.name, self._get_gtk_doc_value(), self.description)
def to_gtk_doc_tag(self):
return '%s: %s%s' % (self.name.capitalize(),
self._get_gtk_doc_value(),
- self.comment or '')
+ self.description or '')
def validate(self):
if self.name == TAG_ATTRIBUTES:
@@ -1074,7 +1074,7 @@ class AnnotationParser(object):
tag = DocTag(comment_block, param_name)
tag.position = position
- tag.comment = param_description
+ tag.description = param_description
if param_annotations:
tag.annotations = self.parse_annotations(tag, param_annotations)
if param_name == TAG_RETURNS:
@@ -1116,10 +1116,10 @@ class AnnotationParser(object):
in_part = PART_DESCRIPTION
part_indent = line_indent
- if not comment_block.comment:
- comment_block.comment = tag_description
+ if not comment_block.description:
+ comment_block.description = tag_description
else:
- comment_block.comment += '\n' + tag_description
+ comment_block.description += '\n' + tag_description
continue
# Now that the deprecated stuff is out of the way, continue parsing real tags
@@ -1145,7 +1145,7 @@ class AnnotationParser(object):
tag = DocTag(comment_block, TAG_RETURNS)
tag.position = position
- tag.comment = tag_description
+ tag.description = tag_description
if tag_annotations:
tag.annotations = self.parse_annotations(tag, tag_annotations)
comment_block.tags[TAG_RETURNS] = tag
@@ -1178,23 +1178,23 @@ class AnnotationParser(object):
# comment block, parameter or tag description.
####################################################################
if in_part in [PART_IDENTIFIER, PART_DESCRIPTION]:
- if not comment_block.comment:
- comment_block.comment = line
+ if not comment_block.description:
+ comment_block.description = line
else:
- comment_block.comment += '\n' + line
+ comment_block.description += '\n' + line
continue
elif in_part == PART_PARAMETERS:
self._validate_multiline_annotation_continuation(line, original_line,
column_offset, position)
# Append to parameter description.
- current_param.comment += ' ' + line.strip()
+ current_param.description += ' ' + line.strip()
continue
elif in_part == PART_TAGS:
self._validate_multiline_annotation_continuation(line, original_line,
column_offset, position)
# Append to tag description.
if current_tag.name.lower() in [TAG_RETURNS, TAG_RETURNVALUE]:
- current_tag.comment += ' ' + line.strip()
+ current_tag.description += ' ' + line.strip()
else:
current_tag.value += ' ' + line.strip()
continue
@@ -1205,8 +1205,8 @@ class AnnotationParser(object):
if comment_block:
# We have picked up a couple of \n characters that where not
# intended. Strip those.
- if comment_block.comment:
- comment_block.comment = comment_block.comment.strip()
+ if comment_block.description:
+ comment_block.description = comment_block.description.strip()
for tag in comment_block.tags.values():
self._clean_comment_block_part(tag)
@@ -1221,10 +1221,10 @@ class AnnotationParser(object):
return None
def _clean_comment_block_part(self, part):
- if part.comment:
- part.comment = part.comment.strip()
+ if part.description:
+ part.description = part.description.strip()
else:
- part.comment = None
+ part.description = None
if part.value:
part.value = part.value.strip()
diff --git a/giscanner/maintransformer.py b/giscanner/maintransformer.py
index aed58b66..a7ca2217 100644
--- a/giscanner/maintransformer.py
+++ b/giscanner/maintransformer.py
@@ -218,7 +218,7 @@ class MainTransformer(object):
section_name = 'SECTION:%s' % (name.lower(), )
block = self._blocks.get(section_name)
if block:
- node.doc = block.comment if block.comment else ''
+ node.doc = block.description if block.description else ''
if isinstance(node, (ast.Class, ast.Interface)):
for prop in node.properties:
self._apply_annotations_property(node, prop)
@@ -577,8 +577,8 @@ class MainTransformer(object):
or node.type.target_giname == 'Gio.Cancellable'):
node.allow_none = True
- if tag is not None and tag.comment is not None:
- node.doc = tag.comment
+ if tag is not None and tag.description is not None:
+ node.doc = tag.description
if ANN_SKIP in annotations:
node.skip = True
@@ -591,7 +591,7 @@ class MainTransformer(object):
if block is None:
return
- node.doc = block.comment if block.comment else ''
+ node.doc = block.description if block.description else ''
since_tag = block.tags.get(TAG_SINCE)
if since_tag is not None:
@@ -813,7 +813,7 @@ class MainTransformer(object):
for m in node.members:
tag = block.params.get(m.symbol, None)
if tag is not None:
- m.doc = tag.comment
+ m.doc = tag.description
def _pass_read_annotations2(self, node, chain):
if isinstance(node, ast.Function):
diff --git a/tests/scanner/annotationparser/test_parser.py b/tests/scanner/annotationparser/test_parser.py
index b20cf2cd..82ca43ba 100644
--- a/tests/scanner/annotationparser/test_parser.py
+++ b/tests/scanner/annotationparser/test_parser.py
@@ -172,13 +172,13 @@ class TestCommentBlock(unittest.TestCase):
parsed += ' </options>\n'
parsed += ' </annotation>\n'
parsed += ' </annotations>\n'
- if param.comment or param.value:
- parsed += ' <description>%s</description>\n' % (param.comment or param.value, )
+ if param.description or param.value:
+ parsed += ' <description>%s</description>\n' % (param.description or param.value, )
parsed += ' </parameter>\n'
parsed += ' </parameters>\n'
- if docblock.comment or docblock.value:
- parsed += ' <description>%s</description>\n' % (docblock.comment or docblock.value, )
+ if docblock.description or docblock.value:
+ parsed += ' <description>%s</description>\n' % (docblock.description or docblock.value, )
if docblock.tags:
parsed += ' <tags>\n'
@@ -203,8 +203,8 @@ class TestCommentBlock(unittest.TestCase):
parsed += ' </options>\n'
parsed += ' </annotation>\n'
parsed += ' </annotations>\n'
- if tag.comment or tag.value:
- parsed += ' <description>%s</description>\n' % (tag.comment or tag.value, )
+ if tag.description or tag.value:
+ parsed += ' <description>%s</description>\n' % (tag.description or tag.value, )
parsed += ' </tag>\n'
parsed += ' </tags>\n'