def wrapped_match_method(self, segments: tuple, parse_context):
"""A wrapper on the match function to do some basic validation."""
# Use the ephemeral_segment if present. This should only
# be the case for grammars where `ephemeral_name` is defined.
if self.ephemeral_name:
# We're going to return as though it's a full match, similar to Anything().
new_grammar = copy.copy(self)
# Reset the ephemeral name on the new version of the grammar otherwise
# we get infinite recursion.
new_grammar.ephemeral_name = None
# We shouldn't allow nested ephemerals. If they're present, don't create
# another. This can happen when grammars call super() on their match method.
if len(segments) == 1 and segments[0].is_type("ephemeral"):
return MatchResult.from_matched(segments)
else:
return MatchResult.from_matched((
EphemeralSegment(
segments=segments,
pos_marker=None,
# Ephemeral segments get a copy of the parent grammar.
parse_grammar=new_grammar,
name=self.ephemeral_name,
), ))
else:
# Otherwise carry on through with wrapping the function.
return func(self, segments, parse_context=parse_context)
def match(self, segments, parse_context):
"""Matches... Anything.
Most useful in match grammars, where a later parse grammar
will work out what's inside.
"""
return MatchResult.from_matched(segments)
def wrapped_match_method(self_cls, segments: tuple, parse_context):
"""A wrapper on the match function to do some basic validation."""
# Use the ephemeral_segment if present. This should only
# be the case for grammars where `ephemeral_name` is defined.
ephemeral_segment = getattr(self_cls, "ephemeral_segment", None)
if ephemeral_segment:
# We're going to return as though it's a full match, similar to Anything().
m = MatchResult.from_matched(
ephemeral_segment(segments=segments))
else:
# Otherwise carry on through with wrapping the function.
m = func(self_cls, segments, parse_context=parse_context)
# Validate result
if not isinstance(m, MatchResult):
parse_context.logger.warning(
"{0}.match, returned {1} rather than MatchResult".format(
func.__qualname__, type(m)))
# Log the result.
WrapParseMatchLogObject(
grammar=func.__qualname__,
func="match",
match=m,
parse_context=parse_context,
segments=segments,
v_level=v_level,
).log()
# Basic Validation, skipped here because it still happens in the parse commands.
return m
def match(self, segments: Tuple[BaseSegment, ...],
parse_context: ParseContext) -> "MatchResult":
"""Matches... Anything.
Most useful in match grammars, where a later parse grammar
will work out what's inside.
"""
return MatchResult.from_matched(segments)
def test__parser__match_add(input_func, raw_seg):
"""Test construction of MatchResults."""
m1 = MatchResult.from_matched([raw_seg])
# Test adding
m2 = m1 + input_func(raw_seg)
# Check it's a match result
assert isinstance(m2, MatchResult)
# In all cases, it should also be of length 2
assert len(m2) == 2
def wrapped_match_method(self, segments: tuple, parse_context):
"""A wrapper on the match function to do some basic validation."""
# Use the ephemeral_segment if present. This should only
# be the case for grammars where `ephemeral_name` is defined.
if self.ephemeral_name:
# We're going to return as though it's a full match, similar to Anything().
new_grammar = copy.copy(self)
# Reset the ephemeral name on the new version of the grammar otherwise
# we get infinite recursion.
new_grammar.ephemeral_name = None
return MatchResult.from_matched((
EphemeralSegment(
segments=segments,
pos_marker=None,
# Ephemeral segments get a copy of the parent grammar.
parse_grammar=new_grammar,
name=self.ephemeral_name,
), ))
else:
# Otherwise carry on through with wrapping the function.
return func(self, segments, parse_context=parse_context)
def match(cls, segments: Tuple["BaseSegment", ...],
parse_context: ParseContext) -> MatchResult:
"""Match a list of segments against this segment.
Note: Match for segments is done in the ABSTRACT.
When dealing with concrete then we're always in parse.
Parse is what happens during expand.
Matching can be done from either the raw or the segments.
This raw function can be overridden, or a grammar defined
on the underlying class.
"""
# Edge case, but it's possible that we have *already matched* on
# a previous cycle. Do should first check whether this is a case
# of that.
if len(segments) == 1 and isinstance(segments[0], cls):
# This has already matched. Winner.
parse_match_logging(
cls.__name__,
"_match",
"SELF",
parse_context=parse_context,
v_level=3,
symbol="+++",
)
return MatchResult.from_matched(segments)
elif len(segments) > 1 and isinstance(segments[0], cls):
parse_match_logging(
cls.__name__,
"_match",
"SELF",
parse_context=parse_context,
v_level=3,
symbol="+++",
)
# This has already matched, but only partially.
return MatchResult((segments[0], ), segments[1:])
if cls.match_grammar:
# Call the private method
with parse_context.deeper_match() as ctx:
m = cls.match_grammar.match(segments=segments,
parse_context=ctx)
# Calling unify here, allows the MatchResult class to do all the type checking.
if not isinstance(m, MatchResult):
raise TypeError(
"[PD:{0} MD:{1}] {2}.match. Result is {3}, not a MatchResult!"
.format(
parse_context.parse_depth,
parse_context.match_depth,
cls.__name__,
type(m),
))
# Once unified we can deal with it just as a MatchResult
if m.has_match():
return MatchResult((cls(segments=m.matched_segments), ),
m.unmatched_segments)
else:
return MatchResult.from_unmatched(segments)
else:
raise NotImplementedError(
"{0} has no match function implemented".format(cls.__name__))
def _longest_trimmed_match(
cls,
segments: Tuple["BaseSegment", ...],
matchers: List["MatchableType"],
parse_context: ParseContext,
trim_noncode=True,
) -> Tuple[MatchResult, Optional["MatchableType"]]:
"""Return longest match from a selection of matchers.
Prioritise the first match, and if multiple match at the same point the longest.
If two matches of the same length match at the same time, then it's the first in
the iterable of matchers.
Returns:
`tuple` of (match_object, matcher).
"""
# Have we been passed an empty list?
if len(segments) == 0:
return MatchResult.from_empty(), None
# If gaps are allowed, trim the ends.
if trim_noncode:
pre_nc, segments, post_nc = trim_non_code_segments(segments)
best_match_length = 0
# iterate at this position across all the matchers
for matcher in matchers:
# MyPy seems to require a type hint here. Not quite sure why.
res_match: MatchResult = matcher.match(
segments, parse_context=parse_context
)
if res_match.is_complete():
# Just return it! (WITH THE RIGHT OTHER STUFF)
if trim_noncode:
return (
MatchResult.from_matched(
pre_nc + res_match.matched_segments + post_nc
),
matcher,
)
else:
return res_match, matcher
elif res_match:
# We've got an incomplete match, if it's the best so far keep it.
if res_match.matched_length > best_match_length:
best_match = res_match, matcher
best_match_length = res_match.matched_length
# If we get here, then there wasn't a complete match. If we
# has a best_match, return that.
if best_match_length > 0:
if trim_noncode:
return (
MatchResult(
pre_nc + best_match[0].matched_segments,
best_match[0].unmatched_segments + post_nc,
),
best_match[1],
)
else:
return best_match
# If no match at all, return nothing
return MatchResult.from_unmatched(segments), None
def match(self, segments: Tuple[BaseSegment, ...],
parse_context: ParseContext) -> MatchResult:
"""Match an arbitrary number of elements separated by a delimiter.
Note that if there are multiple elements passed in that they will be treated
as different options of what can be delimited, rather than a sequence.
"""
# Have we been passed an empty list?
if len(segments) == 0:
return MatchResult.from_empty()
# Make some buffers
seg_buff = segments
matched_segments = MatchResult.from_empty()
# delimiters is a list of tuples containing delimiter segments as we find them.
delimiters: List[BaseSegment] = []
# First iterate through all the segments, looking for the delimiter.
# Second, split the list on each of the delimiters, and ensure that
# each sublist in turn matches one of the elements.
# In more detail, match against delimiter, if we match, put a slice
# up to that point onto a list of slices. Carry on.
while True:
# Check to see whether we've exhausted the buffer, either by iterating through it,
# or by consuming all the non-code segments already.
# NB: If we're here then we've already tried matching the remaining segments against
# the content, so we must be in a trailing case.
if len(seg_buff) == 0:
# Append the remaining buffer in case we're in the not is_code case.
matched_segments += seg_buff
# Nothing left, this is potentially a trailing case?
if self.allow_trailing and (
self.min_delimiters is None
or len(delimiters) >= self.min_delimiters):
# It is! (nothing left so no unmatched segments to append)
return MatchResult.from_matched(
matched_segments.matched_segments)
else:
return MatchResult.from_unmatched(segments)
# We rely on _bracket_sensitive_look_ahead_match to do the bracket counting
# element of this now. We look ahead to find a delimiter or terminator.
matchers = [self.delimiter]
if self.terminator:
matchers.append(self.terminator)
# If gaps aren't allowed, a gap (or non-code segment), acts like a terminator.
if not self.allow_gaps:
matchers.append(NonCodeMatcher())
with parse_context.deeper_match() as ctx:
(
pre_content,
delimiter_match,
delimiter_matcher,
) = self._bracket_sensitive_look_ahead_match(
seg_buff,
matchers,
parse_context=ctx,
)
# Keep track of the *length* of this pre-content section before we start
# to change it later. We need this for dealing with terminators.
pre_content_len = len(pre_content)
# Have we found a delimiter or terminator looking forward?
if delimiter_match:
if delimiter_matcher is self.delimiter:
# Yes. Store it and then match the contents up to now.
delimiters.append(delimiter_match.matched_segments)
# We now test the intervening section as to whether it matches one
# of the things we're looking for. NB: If it's of zero length then
# we return without trying it.
if len(pre_content) > 0:
with parse_context.deeper_match() as ctx:
match, matcher = self._longest_trimmed_match(
segments=pre_content,
matchers=self._elements,
parse_context=ctx,
trim_noncode=self.allow_gaps,
)
# No match, or an incomplete match: Not allowed
if not match or not match.is_complete():
return MatchResult.from_unmatched(segments)
# We have a complete match!
# First add the segment up to the delimiter to the matched segments
matched_segments += match.matched_segments
# Then it depends what we matched.
# Delimiter
if delimiter_matcher is self.delimiter:
# Then add the delimiter to the matched segments
matched_segments += delimiter_match.matched_segments
# Break this for loop and move on, looking for the next delimiter
seg_buff = delimiter_match.unmatched_segments
# Still got some buffer left. Carry on.
continue
# Terminator (or the gap terminator).
elif delimiter_matcher is self.terminator or isinstance(
delimiter_matcher, NonCodeMatcher):
# We just return straight away here. We don't add the terminator to
# this match, it should go with the unmatched parts. The terminator
# may also have mutated the returned segments so we also DON'T want
# the mutated version, it can do that itself (so we return `seg_buff`
# and not `delimiter_match.all_segments()``)
# First check we've had enough delimiters
if (self.min_delimiters
and len(delimiters) < self.min_delimiters):
return MatchResult.from_unmatched(segments)
else:
return MatchResult(
matched_segments.matched_segments,
# Return the part of the seg_buff which isn't in the
# pre-content.
seg_buff[pre_content_len:],
)
else:
raise RuntimeError((
"I don't know how I got here. Matched instead on {0}, which "
"doesn't appear to be delimiter or terminator"
).format(delimiter_matcher))
else:
# Zero length section between delimiters, or zero code
# elements if appropriate. Return unmatched.
return MatchResult.from_unmatched(segments)
else:
# No match for a delimiter looking forward, this means we're
# at the end. In this case we look for a potential partial match
# looking forward. We know it's a non-zero length section because
# we checked that up front.
# First check we're had enough delimiters, because if we haven't then
# there's no sense to try matching
if self.min_delimiters and len(
delimiters) < self.min_delimiters:
return MatchResult.from_unmatched(segments)
# We use the whitespace padded match to hoover up whitespace if enabled,
# and default to the longest matcher. We don't care which one matches.
with parse_context.deeper_match() as ctx:
mat, _ = self._longest_trimmed_match(
seg_buff,
self._elements,
parse_context=ctx,
trim_noncode=self.allow_gaps,
)
if mat:
# We've got something at the end. Return!
if mat.unmatched_segments:
# We have something unmatched and so we should let it also have the trailing elements
return MatchResult(
matched_segments.matched_segments +
mat.matched_segments,
mat.unmatched_segments,
)
else:
# If there's nothing unmatched in the most recent match, then we can consume the trailing
# non code segments
return MatchResult.from_matched(
matched_segments.matched_segments +
mat.matched_segments, )
else:
# No match at the end, are we allowed to trail? If we are then return,
# otherwise we fail because we can't match the last element.
if self.allow_trailing:
return MatchResult(matched_segments.matched_segments,
seg_buff)
else:
return MatchResult.from_unmatched(segments)
def test__parser__match_add_raises(fail_case, raw_seg):
"""Test construction of MatchResults."""
m1 = MatchResult.from_matched([raw_seg])
# Test adding, and check we get an exception of the right type
with pytest.raises(TypeError):
m1 + fail_case
def greedy_match(
cls,
segments,
parse_context,
matchers,
enforce_whitespace_preceeding_terminator,
include_terminator=False,
):
"""Matching for GreedyUntil works just how you'd expect."""
seg_buff = segments
seg_bank = () # Empty tuple
# If no terminators then just return the whole thing.
if matchers == [None]:
return MatchResult.from_matched(segments)
while True:
with parse_context.deeper_match() as ctx:
pre, mat, matcher = cls._bracket_sensitive_look_ahead_match(
seg_buff, matchers, parse_context=ctx)
# Do we have a match?
if mat:
# Do we need to enforce whitespace preceding?
if enforce_whitespace_preceeding_terminator:
# Does the match include some whitespace already?
# Work forward
idx = 0
while True:
elem = mat.matched_segments[idx]
if elem.is_meta:
idx += 1
continue
elif elem.is_type("whitespace", "newline"):
allowable_match = True
break
else:
# No whitespace before. Not allowed.
allowable_match = False
break
# If we're not ok yet, work backward to the preceding sections.
if not allowable_match:
idx = -1
while True:
if len(pre) < abs(idx):
# If we're at the start, it's ok
allowable_match = True
break
if pre[idx].is_meta:
idx -= 1
continue
elif pre[idx].is_type("whitespace", "newline"):
allowable_match = True
break
else:
# No whitespace before. Not allowed.
allowable_match = False
break
# If this match isn't preceded by whitespace and that is
# a requirement, then we can't use it. Carry on...
if not allowable_match:
# Update our buffers and continue onward
seg_bank = pre + mat.matched_segments
seg_buff = mat.unmatched_segments
# Loop around, don't return yet
continue
# Depending on whether we found a terminator or not we treat
# the result slightly differently. If no terminator was found,
# we just use the whole unmatched segment. If we did find one,
# we match up until (but not including [unless self.include_terminator
# is true]) that terminator.
if mat:
# Return everything up to the match unless it's a gap matcher.
if include_terminator:
return MatchResult(
seg_bank + pre + mat.matched_segments,
mat.unmatched_segments,
)
# We can't claim any non-code segments, so we trim them off the end.
leading_nc, pre_seg_mid, trailing_nc = trim_non_code_segments(
seg_bank + pre)
return MatchResult(
leading_nc + pre_seg_mid,
trailing_nc + mat.all_segments(),
)
# No terminator, just return the whole thing.
return MatchResult.from_matched(mat.unmatched_segments)
else:
# Return everything
return MatchResult.from_matched(segments)
def match(
self,
segments: Tuple[BaseSegment, ...],
parse_context: ParseContext,
) -> MatchResult:
"""Match an arbitrary number of elements separated by a delimiter.
Note that if there are multiple elements passed in that they will be treated
as different options of what can be delimited, rather than a sequence.
"""
# Have we been passed an empty list?
if len(segments) == 0:
return MatchResult.from_empty()
# Make some buffers
seg_buff = segments
matched_segments: Tuple[BaseSegment, ...] = ()
unmatched_segments: Tuple[BaseSegment, ...] = ()
cached_matched_segments: Tuple[BaseSegment, ...] = ()
cached_unmatched_segments: Tuple[BaseSegment, ...] = ()
delimiters = 0
matched_delimiter = False
# We want to render progress bar only for the main matching loop,
# so disable it when in deeper parsing.
disable_progress_bar = (
parse_context.parse_depth > 0
or progress_bar_configuration.disable_progress_bar)
# We use amount of `NewLineSegment` to estimate how many steps could be in
# a big file. It's not perfect, but should do a job in most cases.
new_line_segments = [
s for s in segments if isinstance(s, NewlineSegment)
]
progressbar_matching = tqdm(
total=len(new_line_segments),
desc="matching",
miniters=30,
disable=disable_progress_bar,
leave=False,
)
seeking_delimiter = False
has_matched_segs = False
terminated = False
delimiter_matchers = [self.delimiter]
terminator_matchers = []
if self.terminator:
terminator_matchers.append(self.terminator)
# If gaps aren't allowed, a gap (or non-code segment), acts like a terminator.
if not self.allow_gaps:
terminator_matchers.append(NonCodeMatcher())
while True:
progressbar_matching.update(n=1)
if seeking_delimiter:
elements = delimiter_matchers
else:
elements = self._elements
if len(seg_buff) > 0:
pre_non_code, seg_content, post_non_code = trim_non_code_segments(
seg_buff)
if not self.allow_gaps and any(seg.is_whitespace
for seg in pre_non_code):
unmatched_segments = seg_buff
break
if not seg_content: # pragma: no cover
matched_segments += pre_non_code
break
# Check whether there is a terminator before checking for content
with parse_context.deeper_match() as ctx:
match, _ = self._longest_trimmed_match(
segments=seg_content,
matchers=terminator_matchers,
parse_context=ctx,
# We've already trimmed
trim_noncode=False,
)
if match:
terminated = True
unmatched_segments = (pre_non_code +
match.all_segments() +
post_non_code)
break
with parse_context.deeper_match() as ctx:
match, _ = self._longest_trimmed_match(
segments=seg_content,
matchers=elements,
parse_context=ctx,
# We've already trimmed
trim_noncode=False,
terminators=delimiter_matchers
if elements != delimiter_matchers else None,
)
if match:
if elements == delimiter_matchers:
delimiters += 1
matched_delimiter = True
cached_matched_segments = matched_segments
cached_unmatched_segments = seg_buff
else:
matched_delimiter = False
has_matched_segs = True
seg_buff = match.unmatched_segments + post_non_code
unmatched_segments = match.unmatched_segments
if match.is_complete():
matched_segments += (pre_non_code +
match.matched_segments +
post_non_code)
unmatched_segments = match.unmatched_segments
break
matched_segments += pre_non_code + match.matched_segments
seeking_delimiter = not seeking_delimiter
else:
matched_segments += pre_non_code
unmatched_segments = match.unmatched_segments + post_non_code
break
else:
break # pragma: no cover
if self.min_delimiters:
if delimiters < self.min_delimiters:
return MatchResult.from_unmatched(matched_segments +
unmatched_segments)
if terminated:
if has_matched_segs:
return MatchResult(matched_segments, unmatched_segments)
else:
return MatchResult.from_unmatched(matched_segments +
unmatched_segments)
if matched_delimiter and not self.allow_trailing:
if not unmatched_segments:
return MatchResult.from_unmatched(matched_segments +
unmatched_segments)
else:
return MatchResult(cached_matched_segments,
cached_unmatched_segments)
if not has_matched_segs:
return MatchResult.from_unmatched(matched_segments +
unmatched_segments)
if not unmatched_segments:
return MatchResult.from_matched(matched_segments)
return MatchResult(matched_segments, unmatched_segments)
def match(
self,
segments: Tuple[BaseSegment, ...],
parse_context: ParseContext,
) -> MatchResult:
"""Match an arbitrary number of elements separated by a delimiter.
Note that if there are multiple elements passed in that they will be treated
as different options of what can be delimited, rather than a sequence.
"""
# Have we been passed an empty list?
if len(segments) == 0:
return MatchResult.from_empty()
# Make some buffers
seg_buff = segments
matched_segments = MatchResult.from_empty()
# delimiters is a list of tuples containing delimiter segments as we find them.
delimiters: List[BaseSegment] = []
# We want to render progress bar only for the main matching loop,
# so disable it when in deeper parsing.
disable_progress_bar = (
parse_context.parse_depth > 0
or progress_bar_configuration.disable_progress_bar
)
# We use amount of `NewLineSegment` to estimate how many steps could be in
# a big file. It's not perfect, but should do a job in most cases.
new_line_segments = [s for s in segments if isinstance(s, NewlineSegment)]
progressbar_matching = tqdm(
total=len(new_line_segments),
desc="matching",
miniters=30,
disable=disable_progress_bar,
leave=False,
)
# First iterate through all the segments, looking for the delimiter.
# Second, split the list on each of the delimiters, and ensure that
# each sublist in turn matches one of the elements.
# In more detail, match against delimiter, if we match, put a slice
# up to that point onto a list of slices. Carry on.
while True:
progressbar_matching.update(n=1)
# Check to see whether we've exhausted the buffer, either by iterating
# through it, or by consuming all the non-code segments already.
# NB: If we're here then we've already tried matching the remaining segments
# against the content, so we must be in a trailing case.
if len(seg_buff) == 0:
# Append the remaining buffer in case we're in the not is_code case.
matched_segments += seg_buff
# Nothing left, this is potentially a trailing case?
if self.allow_trailing and (
self.min_delimiters is None
or len(delimiters) >= self.min_delimiters
): # pragma: no cover TODO?
# It is! (nothing left so no unmatched segments to append)
return MatchResult.from_matched(matched_segments.matched_segments)
else: # pragma: no cover TODO?
return MatchResult.from_unmatched(segments)
# We rely on _bracket_sensitive_look_ahead_match to do the bracket counting
# element of this now. We look ahead to find a delimiter or terminator.
matchers = [self.delimiter]
if self.terminator:
matchers.append(self.terminator)
# If gaps aren't allowed, a gap (or non-code segment), acts like a
# terminator.
if not self.allow_gaps:
matchers.append(NonCodeMatcher())
with parse_context.deeper_match() as ctx:
(
pre_content,
delimiter_match,
delimiter_matcher,
) = self._bracket_sensitive_look_ahead_match(
seg_buff,
matchers,
parse_context=ctx,
bracket_pairs_set=self.bracket_pairs_set,
)
# Store the mutated segments to reuse.
mutated_segments = pre_content + delimiter_match.all_segments()
# Have we found a delimiter or terminator looking forward?
if delimiter_match:
if delimiter_matcher is self.delimiter:
# Yes. Store it and then match the contents up to now.
delimiters.append(delimiter_match.matched_segments)
# We now test the intervening section as to whether it matches one
# of the things we're looking for. NB: If it's of zero length then
# we return without trying it.
if len(pre_content) > 0:
pre_non_code, pre_content, post_non_code = trim_non_code_segments(
pre_content
)
# Check for whitespace gaps.
# We do this explicitly here rather than relying on an
# untrimmed match so we can handle _whitespace_ explicitly
# compared to other non code segments like placeholders.
if not self.allow_gaps and any(
seg.is_whitespace for seg in pre_non_code + post_non_code
):
return MatchResult.from_unmatched(
mutated_segments
) # pragma: no cover TODO?
with parse_context.deeper_match() as ctx:
match, _ = self._longest_trimmed_match(
segments=pre_content,
matchers=self._elements,
parse_context=ctx,
# We've already trimmed
trim_noncode=False,
)
# No match - Not allowed
if not match:
if self.allow_trailing:
# If we reach this point, the lookahead match has hit a
# delimiter beyond the scope of this Delimited section.
# Trailing delimiters are allowed, so return matched up to
# this section.
return MatchResult(
matched_segments.matched_segments,
pre_non_code
+ match.unmatched_segments
+ post_non_code
+ delimiter_match.all_segments(),
)
else:
return MatchResult.from_unmatched(mutated_segments)
if not match.is_complete():
# If we reach this point, the lookahead match has hit a
# delimiter beyond the scope of this Delimited section. We
# should return a partial match, and the delimiter as unmatched.
return MatchResult(
matched_segments.matched_segments
+ pre_non_code
+ match.matched_segments,
match.unmatched_segments
+ post_non_code
+ delimiter_match.all_segments(),
)
# We have a complete match!
# First add the segment up to the delimiter to the matched segments
matched_segments += (
pre_non_code + match.matched_segments + post_non_code
)
# Then it depends what we matched.
# Delimiter
if delimiter_matcher is self.delimiter:
# Then add the delimiter to the matched segments
matched_segments += delimiter_match.matched_segments
# Break this for loop and move on, looking for the next
# delimiter
seg_buff = delimiter_match.unmatched_segments
# Still got some buffer left. Carry on.
continue
# Terminator (or the gap terminator).
elif delimiter_matcher is self.terminator or isinstance(
delimiter_matcher, NonCodeMatcher
):
# We just return straight away here. We don't add the terminator
# to this match, it should go with the unmatched parts.
# First check we've had enough delimiters
if (
self.min_delimiters
and len(delimiters) < self.min_delimiters
):
return MatchResult.from_unmatched(mutated_segments)
else:
return MatchResult(
matched_segments.matched_segments,
delimiter_match.all_segments(),
)
else: # pragma: no cover
raise RuntimeError(
(
"I don't know how I got here. Matched instead on {}, "
"which doesn't appear to be delimiter or terminator"
).format(delimiter_matcher)
)
else:
# Zero length section between delimiters, or zero code
# elements if appropriate. Return unmatched.
return MatchResult.from_unmatched(mutated_segments)
else:
# No match for a delimiter looking forward, this means we're
# at the end. In this case we look for a potential partial match
# looking forward. We know it's a non-zero length section because
# we checked that up front.
# First check we're had enough delimiters, because if we haven't then
# there's no sense to try matching
if self.min_delimiters and len(delimiters) < self.min_delimiters:
return MatchResult.from_unmatched(mutated_segments)
# We use the whitespace padded match to hoover up whitespace if enabled,
# and default to the longest matcher. We don't care which one matches.
pre_non_code, trimmed_segments, post_non_code = trim_non_code_segments(
mutated_segments
)
# Check for whitespace gaps.
# We do this explicitly here rather than relying on an
# untrimmed match so we can handle _whitespace_ explicitly
# compared to other non code segments like placeholders.
if not self.allow_gaps and any(
seg.is_whitespace for seg in pre_non_code + post_non_code
):
return MatchResult.from_unmatched(
mutated_segments
) # pragma: no cover TODO?
with parse_context.deeper_match() as ctx:
mat, _ = self._longest_trimmed_match(
trimmed_segments,
self._elements,
parse_context=ctx,
# We've already trimmed
trim_noncode=False,
)
if mat:
# We've got something at the end. Return!
if mat.unmatched_segments:
# We have something unmatched and so we should let it also have
# the trailing elements
return MatchResult(
matched_segments.matched_segments
+ pre_non_code
+ mat.matched_segments,
mat.unmatched_segments + post_non_code,
)
else:
# If there's nothing unmatched in the most recent match, then we
# can consume the trailing non code segments
return MatchResult.from_matched(
matched_segments.matched_segments
+ pre_non_code
+ mat.matched_segments
+ post_non_code,
)
else:
# No match at the end, are we allowed to trail? If we are then
# return, otherwise we fail because we can't match the last element.
if self.allow_trailing:
return MatchResult(matched_segments.matched_segments, seg_buff)
else:
return MatchResult.from_unmatched(mutated_segments)
def _longest_trimmed_match(
cls,
segments: Tuple[BaseSegment, ...],
matchers: List[MatchableType],
parse_context: ParseContext,
trim_noncode=True,
terminators: List[MatchableType] = None,
) -> Tuple[MatchResult, Optional[MatchableType]]:
"""Return longest match from a selection of matchers.
Prioritise the first match, and if multiple match at the same point the longest.
If two matches of the same length match at the same time, then it's the first in
the iterable of matchers.
Returns:
`tuple` of (match_object, matcher).
"""
terminated = False
# Have we been passed an empty list?
if len(segments) == 0: # pragma: no cover
return MatchResult.from_empty(), None
# If gaps are allowed, trim the ends.
if trim_noncode:
pre_nc, segments, post_nc = trim_non_code_segments(segments)
best_match_length = 0
# iterate at this position across all the matchers
for matcher in matchers:
# MyPy seems to require a type hint here. Not quite sure why.
res_match: MatchResult = matcher.match(segments,
parse_context=parse_context)
if res_match.is_complete():
# Just return it! (WITH THE RIGHT OTHER STUFF)
if trim_noncode:
return (
MatchResult.from_matched(pre_nc +
res_match.matched_segments +
post_nc),
matcher,
)
else:
return res_match, matcher
elif res_match:
# We've got an incomplete match, if it's the best so far keep it.
if res_match.trimmed_matched_length > best_match_length:
best_match = res_match, matcher
best_match_length = res_match.trimmed_matched_length
if terminators:
_, segs, _ = trim_non_code_segments(
best_match[0].unmatched_segments)
for terminator in terminators:
terminator_match: MatchResult = terminator.match(
segs, parse_context=parse_context)
if terminator_match.matched_segments:
terminated = True
break
if terminated:
break
# We could stash segments here, but given we might have some successful
# matches here, we shouldn't, because they'll be mutated in the wrong way.
# Eventually there might be a performance gain from doing that sensibly
# here.
# If we get here, then there wasn't a complete match. If we
# has a best_match, return that.
if best_match_length > 0:
if trim_noncode:
return (
MatchResult(
pre_nc + best_match[0].matched_segments,
best_match[0].unmatched_segments + post_nc,
),
best_match[1],
)
else:
return best_match
# If no match at all, return nothing
return MatchResult.from_unmatched(segments), None