Fixes 4969 - Add parsing of filters

abp/filters/parser.py

Index: abp/filters/parser.py
===================================================================
--- a/abp/filters/parser.py
+++ b/abp/filters/parser.py
@@ -13,32 +13,50 @@
 # You should have received a copy of the GNU General Public License
 # along with Adblock Plus.  If not, see <http://www.gnu.org/licenses/>.
 
 from __future__ import unicode_literals
 
 import re
 from collections import namedtuple
 
-__all__ = ['parse_filterlist', 'parse_line', 'ParseError']
+__all__ = ['parse_filterlist', 'parse_line', 'ParseError', 'ST', 'FA']
 
 
 class ParseError(Exception):
     """Exception thrown by the parser when it encounters invalid input.
 
     :param error: Description of the error.
     :param text: The text which was being parsed when an error occurred.
     """
 
     def __init__(self, error, text):
         Exception.__init__(self, '{} in "{}"'.format(error, text))
         self.text = text
         self.error = error
 
 
+# Constants related to filters (see https://adblockplus.org/filters).
+class ST:

[mathias, 2017/07/28 16:43:29]

Why abbreviating here (ST) and below (FA)?

[Vasily Kuznetsov, 2017/07/28 17:38:10]

To be completely honest, the reason is kind of stupid: to make flake8 happy :/

This class was called SELECTOR_TYPE originally and the following one
FILTER_ACTION. So then you'd end up with constants like SELECTOR_TYPE.CSS and
FILTER_ACTION.ALLOW. And then if you (as a user of the library) like shortness,
you can import them `from abp.filters import SELECTOR_TYPE as ST` and get your
short ST.CSS. But flake8 doesn't like class names that are not CamelCased. I
thought about rolling a small special class to hold constants, kind of like:

SELECTOR_TYPE = Namespace(
    URL_PATTERN='url-pattern',
    URL_REGEXP='url-regexp',
    ...
)

But then you need to implement this `Namespace` class (or it can be a function),
which is annoying and inelegant. Another option is to just disable N801 in
flake8, that could probably also do it.

[Vasily Kuznetsov, 2017/07/28 18:57:49]

Done.

+    """Selector types"""
+    URL_PATTERN = 'url-pattern'  # Normal URL patterns.
+    URL_REGEXP = 'url-regexp'    # Regular expressions for URLs.
+    CSS = 'css'                  # CSS selectors for hiding filters.
+    XCSS = 'extended-css'        # Extended CSS selectors (to emulate CSS4).
+    ABP_SIMPLE = 'abp-simple'    # Simplified element hiding syntax.
+
+
+class FA:
+    """Filter actions"""
+    BLOCK = 'block'              # Block the request.
+    ALLOW = 'allow'              # Allow the request (whitelist).
+    HIDE = 'hide'                # Hide selected element(s).
+    SHOW = 'show'                # Show selected element(s) (whitelist).
+
+
 def _line_type(name, field_names, format_string):
     """Define a line type.
 
     :param name: The name of the line type to define.
     :param field_names: A sequence of field names or one space-separated
         string that contains all field names.
     :param format_string: A format specifier for converting this line type
         back to string representation.
@@ -51,25 +69,32 @@
     lt.to_string = lambda self: format_string.format(self)
     return lt
 
 
 Header = _line_type('Header', 'version', '[{.version}]')
 EmptyLine = _line_type('EmptyLine', '', '')
 Comment = _line_type('Comment', 'text', '! {.text}')
 Metadata = _line_type('Metadata', 'key value', '! {0.key}: {0.value}')
-Filter = _line_type('Filter', 'expression', '{.expression}')
+Filter = _line_type('Filter', 'text selector action options', '{.text}')
 Include = _line_type('Include', 'target', '%include {0.target}%')
 
 
 METADATA_REGEXP = re.compile(r'!\s*(\w+)\s*:\s*(.*)')
 METADATA_KEYS = {'Homepage', 'Title', 'Expires', 'Checksum', 'Redirect',
                  'Version'}
 INCLUDE_REGEXP = re.compile(r'%include\s+(.+)%')
 HEADER_REGEXP = re.compile(r'\[(Adblock(?:\s*Plus\s*[\d\.]+?)?)\]', flags=re.I)
+HFILTER_REGEXP = re.compile(r'^([^/*|@"!]*?)#([@?])?#(.+)$')

[mathias, 2017/07/28 16:43:29]

Why abbreviating? What's wrong about HIDING_FILTER_REGEXP here and
BLOCKING_FILTER_REGEXP below?

[Vasily Kuznetsov, 2017/07/28 17:38:10]

It's shorter this way. But I don't feel very strong about it, will rename them.

[Vasily Kuznetsov, 2017/07/28 18:57:49]

Done.

+BFILTER_REGEXP_REGEXP = re.compile(

[mathias, 2017/07/28 16:43:30]

I was wondering about the *_REGEXP_REGEXP name, but then where is this actually
used anyway?

[Vasily Kuznetsov, 2017/07/28 17:38:10]

It's the regular expression for blocking filters which use regular expressions,
so the name is legit. However, you're right that I'm not actually using it. Just
copied it from ABP source. The funny thing, it seems that there it's not used
either :) (see
https://hg.adblockplus.org/adblockpluscore/file/tip/lib/filterClasses.js)

I will remove it.

[Vasily Kuznetsov, 2017/07/28 18:57:49]

Done.

+    r'^(@@)?\/.*\/(?:\$~?[\w-]+(?:=[^,\s]+)?(?:,~?[\w-]+(?:=[^,\s]+)?)*)?$'
+)
+BFILTER_OPTIONS_REGEXP = re.compile(
+    r'\$(~?[\w-]+(?:=[^,\s]+)?(?:,~?[\w-]+(?:=[^,\s]+)?)*)$'
+)
 
 
 def _parse_comment(text):
     match = METADATA_REGEXP.match(text)
     if match and match.group(1) in METADATA_KEYS:
         return Metadata(match.group(1), match.group(2))
     return Comment(text[1:].strip())
 
@@ -83,16 +108,92 @@
 
 def _parse_instruction(text):
     match = INCLUDE_REGEXP.match(text)
     if not match:
         raise ParseError('Unrecognized instruction', text)
     return Include(match.group(1))
 
 
+def _parse_option(option):
+    if '=' in option:
+        name, value = option.split('=', 1)
+    elif option.startswith('~'):
+        name, value = option[1:], False
+    else:
+        name, value = option, True
+
+    # Handle special cases of multivalued options.
+    if name == 'domain':

[mathias, 2017/07/28 16:43:30]

Wouldn't it make sense to enumerate recognized OPTION_$NAME symbols as well?

[Vasily Kuznetsov, 2017/07/28 17:38:10]

Yeah, probably makes sense to make some kind of enum for these things. It will
be needed anyway later, for interpretation, and might also be useful for the
users of the library.

[Vasily Kuznetsov, 2017/07/28 18:57:49]

Done.

+        name, value = 'domains', _parse_options(value, '|')

[mathias, 2017/07/28 16:43:30]

Why using a different / plural key for the parsed version?

[Vasily Kuznetsov, 2017/07/28 17:38:10]

Because semantically it's a list, always, so calling it 'domain' is a bit
confusing. I see your point, however, that it might be not very obvious if the
option is named differently than what it's called in the unparsed filter. 

Perhaps this could be solved in a mutually beneficial way if the key will be
'domain' but the constant will be called `OPTION.DOMAINS`. Although this is also
confusing :/

[Vasily Kuznetsov, 2017/07/28 18:57:49]

Done.

+    elif name == 'sitekey':
+        name, value = 'sitekeys', value.split('|')
+
+    return name, value
+
+
+def _parse_options(options, separator=','):
+    return [_parse_option(o) for o in options.split(separator)]
+
+
+def _parse_blocking_filter(text):
+    # Based on RegExpFilter.fromText in lib/filterClasses.js
+    # in https://hg.adblockplus.org/adblockpluscore.
+    action = FA.BLOCK
+    options = []
+    selector = text
+
+    if selector.startswith('@@'):
+        action = FA.ALLOW
+        selector = selector[2:]
+
+    if '$' in selector:
+        opt_match = BFILTER_OPTIONS_REGEXP.search(selector)
+        if opt_match:
+            selector = selector[:opt_match.start(0)]
+            options = _parse_options(opt_match.group(1))
+
+    if (len(selector) > 1 and
+            selector.startswith('/') and selector.endswith('/')):
+        selector = {'type': ST.URL_REGEXP, 'value': selector[1:-1]}
+    else:
+        selector = {'type': ST.URL_PATTERN, 'value': selector}
+
+    return Filter(text, selector, action, options)
+
+
+def _parse_hiding_filter(text, domains, type_flag, selector_value):
+    selector = {'type': ST.CSS, 'value': selector_value}
+    action = FA.HIDE
+    options = []
+
+    if type_flag == '@':
+        action = FA.SHOW
+    elif type_flag == '?':
+        selector['type'] = ST.XCSS
+
+    if domains:
+        options.append(('domains', _parse_options(domains)))
+
+    return Filter(text, selector, action, options)
+
+
+def parse_filter(text):
+    """Parse one filter.
+
+    :param text: Text representation of a filter.
+    :returns: filter object.

[mathias, 2017/07/28 16:43:29]

I think this should be upper-case "Filter".

[Vasily Kuznetsov, 2017/07/28 17:38:10]

Yes.

+    """
+    if '#' in text:
+        match = HFILTER_REGEXP.search(text)
+        if match:
+            return _parse_hiding_filter(text, *match.groups())
+    return _parse_blocking_filter(text)
+
+
 def parse_line(line_text):
     """Parse one line of a filter list.
 
     :param line_text: Line of a filter list (must be a unicode string).
     :returns: Parsed line object (see `_line_type`).
     :raises ParseError: If the line can't be successfully parsed.
     """
     content = line_text.strip()
@@ -101,17 +202,17 @@
         line = EmptyLine()
     elif content.startswith('!'):
         line = _parse_comment(content)
     elif content.startswith('%') and content.endswith('%'):
         line = _parse_instruction(content)
     elif content.startswith('[') and content.endswith(']'):
         line = _parse_header(content)
     else:
-        line = Filter(content)
+        line = parse_filter(content)
 
     assert line.to_string().replace(' ', '') == content.replace(' ', '')
     return line
 
 
 def parse_filterlist(lines):
     """Parse filter list from an iterable.