def test_singleton(self):
"""Test to ensure mutiple calls to benchit return the same object
"""
b = BenchIt()
b2 = BenchIt()
self.assertEqual(b, b2)
self.assertEqual(b.timer_name, "BenchIt")
b3 = BenchIt("New Bencher")
self.assertNotEqual(b, b3)
self.assertEqual(b3.timer_name, "New Bencher")
b4 = BenchIt(timer_name="New Bencher")
def test_benchit(self):
"""
Full benchmark test.
Test starting, marking, stopping, and displaying the benchmark.
"""
b = BenchIt()
# Ensure start key is set on init
self.assertEqual(len(b), 1)
self.assertIn('_start', b.keys())
# Set a marker
b.mark('Unique Marker')
self.assertIn('Unique Marker', b.keys())
# Stop the benchmark
b.stop()
self.assertIn('_end', b.keys())
# Ensure end time is after the start time
self.assertGreater(b['_end'], b['_start'])
# Check the output table
with capture(b.display) as output:
self.assertIn("Avg Time", output)
self.assertIn("Unique Marker", output)
self.assertIn("Total runtime", output)
self.assertIn("test_benchit", output)
def demo_benchmark():
b = BenchIt() # starts the timer
# do stuff
sleep(1); b()
sleep(.2)
sleep(.3)
sleep(.4)
b("More stuff Done")
for i in range(1, 5):
sleep(.1); b()
b.display()
def test_benchit(self):
"""
Full benchmark test.
Test starting, marking, stopping, and displaying the benchmark.
"""
b = BenchIt()
# Ensure start key is set on init
self.assertEqual(len(b.times), 1)
self.assertIn('_start', b.times.keys())
# Set a marker
"code marker"
b()
"more code"
b("__call__ marker")
b.mark('Unique Marker')
self.assertIn('"code marker"', b.times.keys())
self.assertIn('__call__ marker', b.times.keys())
self.assertIn('Unique Marker', b.times.keys())
# Stop the benchmark
b.stop()
self.assertIn('_end', b.times.keys())
# Ensure end time is after the start time
self.assertGreater(b.times['_end'], b.times['_start'])
# Check the output table
with capture(b.display) as output:
self.assertIn("Avg Time", output)
self.assertIn("Unique Marker", output)
self.assertIn("Total runtime", output)
self.assertIn("test_benchit", output)
def parse(self, parse_context=None, parse_grammar=None):
"""Use the parse grammar to find subsegments within this segment.
A large chunk of the logic around this can be found in the `expand` method.
Use the parse setting in the context for testing, mostly to check how deep to go.
True/False for yes or no, an integer allows a certain number of levels.
Optionally, this method allows a custom parse grammar to be
provided which will override any existing parse grammar
on the segment.
"""
# Clear the blacklist cache so avoid missteps
if parse_context:
parse_context.blacklist.clear()
# the parse_depth and recurse kwargs control how deep we will recurse for testing.
if not self.segments:
# This means we're a root segment, just return an unmutated self
return self
# Check the Parse Grammar
parse_grammar = parse_grammar or self.parse_grammar
if parse_grammar is None:
# No parse grammar, go straight to expansion
parse_context.logger.debug(
"{0}.parse: no grammar. Going straight to expansion".format(
self.__class__.__name__))
else:
# For debugging purposes. Ensure that we don't have non-code elements
# at the start or end of the segments. They should always in the middle,
# or in the parent expression.
segments = self.segments
if self.can_start_end_non_code:
pre_nc, segments, post_nc = trim_non_code_segments(segments)
else:
pre_nc = ()
post_nc = ()
if (not segments[0].is_code) and (not segments[0].is_meta):
raise ValueError(
"Segment {0} starts with non code segment: {1!r}.\n{2!r}"
.format(self, segments[0].raw, segments))
if (not segments[-1].is_code) and (not segments[-1].is_meta):
raise ValueError(
"Segment {0} ends with non code segment: {1!r}.\n{2!r}"
.format(self, segments[-1].raw, segments))
# NOTE: No match_depth kwarg, because this is the start of the matching.
with parse_context.matching_segment(
self.__class__.__name__) as ctx:
m = parse_grammar.match(segments=segments, parse_context=ctx)
if not isinstance(m, MatchResult):
raise TypeError(
"[PD:{0}] {1}.match. Result is {2}, not a MatchResult!".
format(parse_context.parse_depth, self.__class__.__name__,
type(m)))
# Basic Validation, that we haven't dropped anything.
check_still_complete(segments, m.matched_segments,
m.unmatched_segments)
if m.has_match():
if m.is_complete():
# Complete match, happy days!
self.segments = pre_nc + m.matched_segments + post_nc
else:
# Incomplete match.
# For now this means the parsing has failed. Lets add the unmatched bit at the
# end as something unparsable.
# TODO: Do something more intelligent here.
self.segments = (
pre_nc + m.matched_segments + (UnparsableSegment(
segments=m.unmatched_segments + post_nc,
expected="Nothing...",
), ))
elif self.allow_empty and not segments:
# Very edge case, but some segments are allowed to be empty other than non-code
self.segments = pre_nc + post_nc
else:
# If there's no match at this stage, then it's unparsable. That's
# a problem at this stage so wrap it in an unparsable segment and carry on.
self.segments = (
pre_nc + (
UnparsableSegment(
segments=segments,
expected=self.name,
), # NB: tuple
) + post_nc)
bencher = BenchIt() # starts the timer
bencher("Parse complete of {0!r}".format(self.__class__.__name__))
# Recurse if allowed (using the expand method to deal with the expansion)
parse_context.logger.debug(
"{0}.parse: Done Parse. Plotting Recursion. Recurse={1!r}".format(
self.__class__.__name__, parse_context.recurse))
parse_depth_msg = "###\n#\n# Beginning Parse Depth {0}: {1}\n#\n###\nInitial Structure:\n{2}".format(
parse_context.parse_depth + 1, self.__class__.__name__,
self.stringify())
if parse_context.may_recurse():
parse_context.logger.debug(parse_depth_msg)
with parse_context.deeper_parse() as ctx:
self.segments = self.expand(self.segments, parse_context=ctx)
return self
def fix(force,
paths,
parallel,
bench=False,
fixed_suffix="",
logger=None,
**kwargs):
"""Fix SQL files.
PATH is the path to a sql file or directory to lint. This can be either a
file ('path/to/file.sql'), a path ('directory/of/sql/files'), a single ('-')
character to indicate reading from *stdin* or a dot/blank ('.'/' ') which will
be interpreted like passing the current working directory as a path argument.
"""
# some quick checks
fixing_stdin = ("-", ) == paths
c = get_config(**kwargs)
lnt, formatter = get_linter_and_formatter(c, silent=fixing_stdin)
verbose = c.get("verbose")
bencher = BenchIt()
formatter.dispatch_config(lnt)
# Set up logging.
set_logging_level(verbosity=verbose,
logger=logger,
stderr_output=fixing_stdin)
# handle stdin case. should output formatted sql to stdout and nothing else.
if fixing_stdin:
stdin = sys.stdin.read()
# TODO: Remove verbose
result = lnt.lint_string_wrapped(stdin, fname="stdin", fix=True)
stdout = result.paths[0].files[0].fix_string()[0]
click.echo(stdout, nl=False)
sys.exit()
# Lint the paths (not with the fix argument at this stage), outputting as we go.
click.echo("==== finding fixable violations ====")
try:
result = lnt.lint_paths(paths,
fix=True,
ignore_non_existent_files=False,
parallel=parallel)
except IOError:
click.echo(
colorize(
"The path(s) {0!r} could not be accessed. Check it/they exist(s)."
.format(paths),
"red",
))
sys.exit(1)
# NB: We filter to linting violations here, because they're
# the only ones which can be potentially fixed.
if result.num_violations(types=SQLLintError, fixable=True) > 0:
click.echo("==== fixing violations ====")
click.echo("{0} fixable linting violations found".format(
result.num_violations(types=SQLLintError, fixable=True)))
if force:
click.echo(colorize("FORCE MODE", "red") + ": Attempting fixes...")
# TODO: Remove verbose
success = do_fixes(
lnt,
result,
formatter,
types=SQLLintError,
fixed_file_suffix=fixed_suffix,
)
if not success:
sys.exit(1)
else:
click.echo("Are you sure you wish to attempt to fix these? [Y/n] ",
nl=False)
c = click.getchar().lower()
click.echo("...")
if c in ("y", "\r", "\n"):
click.echo("Attempting fixes...")
# TODO: Remove verbose
success = do_fixes(
lnt,
result,
formatter,
types=SQLLintError,
fixed_file_suffix=fixed_suffix,
)
if not success:
sys.exit(1)
else:
click.echo("All Finished 📜 🎉!")
elif c == "n":
click.echo("Aborting...")
else:
click.echo("Invalid input, please enter 'Y' or 'N'")
click.echo("Aborting...")
else:
click.echo("==== no fixable linting violations found ====")
if result.num_violations(types=SQLLintError, fixable=False) > 0:
click.echo(" [{0} unfixable linting violations found]".format(
result.num_violations(types=SQLLintError, fixable=False)))
click.echo("All Finished 📜 🎉!")
if bench:
click.echo("\n\n==== bencher stats ====")
bencher.display()
sys.exit(0)
def parse(path, code_only, format, profiler, bench, **kwargs):
"""Parse SQL files and just spit out the result.
PATH is the path to a sql file or directory to lint. This can be either a
file ('path/to/file.sql'), a path ('directory/of/sql/files'), a single ('-')
character to indicate reading from *stdin* or a dot/blank ('.'/' ') which will
be interpreted like passing the current working directory as a path argument.
"""
# Initialise the benchmarker
bencher = BenchIt() # starts the timer
c = get_config(**kwargs)
# We don't want anything else to be logged if we want a yaml output
lnt = get_linter(c, silent=format in ('json', 'yaml'))
verbose = c.get('verbose')
recurse = c.get('recurse')
config_string = format_config(lnt, verbose=verbose)
if len(config_string) > 0:
lnt.log(config_string)
# TODO: do this better
nv = 0
if profiler:
# Set up the profiler if required
try:
import cProfile
except ImportError:
lnt.log('The cProfiler is not available on your platform.')
sys.exit(1)
pr = cProfile.Profile()
pr.enable()
bencher("Parse setup")
try:
# handle stdin if specified via lone '-'
if '-' == path:
# put the parser result in a list to iterate later
config = lnt.config.make_child_from_path('stdin')
result = [lnt.parse_string(
sys.stdin.read(),
'stdin',
verbosity=verbose,
recurse=recurse,
config=config
)]
else:
# A single path must be specified for this command
result = lnt.parse_path(path, verbosity=verbose, recurse=recurse)
# iterative print for human readout
if format == 'human':
for parsed, violations, time_dict in result:
if parsed:
lnt.log(parsed.stringify(code_only=code_only))
else:
# TODO: Make this prettier
lnt.log('...Failed to Parse...')
nv += len(violations)
for v in violations:
lnt.log(format_violation(v, verbose=verbose))
if verbose >= 2:
lnt.log("==== timings ====")
lnt.log(cli_table(time_dict.items()))
bencher("Output details for file")
else:
# collect result and print as single payload
# will need to zip in the file paths
filepaths = ['stdin'] if '-' == path else lnt.paths_from_path(path)
result = [
dict(
filepath=filepath,
segments=parsed.as_record(code_only=code_only, show_raw=True)
)
for filepath, (parsed, _, _) in zip(filepaths, result)
]
if format == 'yaml':
# For yaml dumping always dump double quoted strings if they contain tabs or newlines.
def quoted_presenter(dumper, data):
"""Representer which always double quotes string values needing escapes."""
if '\n' in data or '\t' in data:
return dumper.represent_scalar('tag:yaml.org,2002:str', data, style='"')
else:
return dumper.represent_scalar('tag:yaml.org,2002:str', data, style='')
yaml.add_representer(str, quoted_presenter)
click.echo(yaml.dump(result))
elif format == 'json':
click.echo(json.dumps(result))
except IOError:
click.echo(colorize('The path {0!r} could not be accessed. Check it exists.'.format(path), 'red'))
sys.exit(1)
if profiler:
pr.disable()
profiler_buffer = StringIO()
ps = pstats.Stats(
pr, stream=profiler_buffer
).sort_stats('cumulative')
ps.print_stats()
lnt.log("==== profiler stats ====")
# Only print the first 50 lines of it
lnt.log('\n'.join(profiler_buffer.getvalue().split('\n')[:50]))
if bench:
lnt.log("\n\n==== bencher stats ====")
bencher.display()
if nv > 0:
sys.exit(66)
else:
sys.exit(0)
def parse(self, parse_context=None):
"""Use the parse grammar to find subsegments within this segment.
A large chunk of the logic around this can be found in the `expand` method.
Use the parse setting in the context for testing, mostly to check how deep to go.
True/False for yes or no, an integer allows a certain number of levels.
"""
if not parse_context.dialect:
raise RuntimeError("No dialect provided to {0!r}!".format(self))
# Clear the blacklist cache so avoid missteps
if parse_context:
parse_context.blacklist.clear()
# the parse_depth and recurse kwargs control how deep we will recurse for testing.
if not self.segments:
# This means we're a root segment, just return an unmutated self
return self
# Get the Parse Grammar
g = self._parse_grammar()
if g is None:
# No parse grammar, go straight to expansion
logging.debug("{0}.parse: no grammar. Going straight to expansion".format(self.__class__.__name__))
else:
# Use the Parse Grammar (and the private method)
# For debugging purposes. Ensure that we don't have non-code elements
# at the start or end of the segments. They should always in the middle,
# or in the parent expression.
if not self._can_start_end_non_code:
if (not self.segments[0].is_code) and (not self.segments[0].is_meta):
raise ValueError("Segment {0} starts with non code segment: {1!r}.\n{2!r}".format(
self, self.segments[0].raw, self.segments))
if (not self.segments[-1].is_code) and (not self.segments[-1].is_meta):
raise ValueError("Segment {0} ends with non code segment: {1!r}.\n{2!r}".format(
self, self.segments[-1].raw, self.segments))
# NOTE: No match_depth kwarg, because this is the start of the matching.
m = g._match(
segments=self.segments,
parse_context=parse_context.copy(
match_segment=self.__class__.__name__
)
)
if not isinstance(m, MatchResult):
raise TypeError(
"[PD:{0}] {1}.match. Result is {2}, not a MatchResult!".format(
parse_context.parse_depth, self.__class__.__name__, type(m)))
# Basic Validation, that we haven't dropped anything.
check_still_complete(self.segments, m.matched_segments, m.unmatched_segments)
if m.has_match():
if m.is_complete():
# Complete match, happy days!
self.segments = m.matched_segments
else:
# Incomplete match.
# For now this means the parsing has failed. Lets add the unmatched bit at the
# end as something unparsable.
# TODO: Do something more intelligent here.
self.segments = m.matched_segments + (UnparsableSegment(
segments=m.unmatched_segments, expected="Nothing..."),)
else:
# If there's no match at this stage, then it's unparsable. That's
# a problem at this stage so wrap it in an unparable segment and carry on.
self.segments = (UnparsableSegment(
segments=self.segments,
expected=g.expected_string(dialect=parse_context.dialect)),) # NB: tuple
# Validate new segments
self.validate_segments(text="parsing")
bencher = BenchIt() # starts the timer
bencher("Parse complete of {0!r}".format(self.__class__.__name__))
# Recurse if allowed (using the expand method to deal with the expansion)
logging.debug(
"{0}.parse: Done Parse. Plotting Recursion. Recurse={1!r}".format(
self.__class__.__name__, parse_context.recurse))
parse_depth_msg = "###\n#\n# Beginning Parse Depth {0}: {1}\n#\n###\nInitial Structure:\n{2}".format(
parse_context.parse_depth + 1, self.__class__.__name__, self.stringify())
if parse_context.recurse is True:
logging.debug(parse_depth_msg)
self.segments = self.expand(
self.segments,
parse_context=parse_context.copy(
incr='parse_depth', match_depth=0, recurse=True
)
)
elif isinstance(parse_context.recurse, int):
if parse_context.recurse > 1:
logging.debug(parse_depth_msg)
self.segments = self.expand(
self.segments,
parse_context=parse_context.copy(decr='recurse', incr='parse_depth')
)
# Validate new segments
self.validate_segments(text="expanding")
return self
def parse_string(self, s, fname=None, verbosity=0, recurse=True, config=None):
"""Parse a string.
Returns:
`tuple` of (`parsed`, `violations`, `time_dict`, `config_diff`).
`parsed` is a segment structure representing the parsed file. If
parsing fails due to an inrecoverable violation then we will
return None.
`violations` is a :obj:`list` of violations so far, which will either be
templating, lexing or parsing violations at this stage.
`time_dict` is a :obj:`dict` containing timings for how long each step
took in the process.
"""
violations = []
t0 = time.monotonic()
bencher = BenchIt() # starts the timer
if fname:
short_fname = fname.replace('\\', '/').split('/')[-1]
else:
# this handles to potential case of a null fname
short_fname = fname
bencher("Staring parse_string for {0!r}".format(short_fname))
# Log the start of this process if we're in a more verbose mode.
if verbosity > 1:
self.log(format_filename(filename=fname, success='PARSING', verbose=verbosity))
# This is where we output config diffs if they exist.
if config:
# Only output config diffs if there is a config to diff to.
config_diff = config.diff_to(self.config)
if config_diff:
self.log(" Config Diff:")
self.log(format_config_vals(self.config.iter_vals(cfg=config_diff)))
verbosity_logger("TEMPLATING RAW [{0}] ({1})".format(self.templater.name, fname), verbosity=verbosity)
try:
s = self.templater.process(s, fname=fname, config=config or self.config)
except SQLTemplaterError as err:
violations.append(err)
file_segment = None
# NB: We'll carry on if we fail to template, it might still lex
t1 = time.monotonic()
bencher("Templating {0!r}".format(short_fname))
if s:
verbosity_logger("LEXING RAW ({0})".format(fname), verbosity=verbosity)
# Lex the file and log any problems
try:
file_segment, lex_vs = FileSegment.from_raw(s, config=config or self.config)
# We might just get the violations as a list
violations += lex_vs
except SQLLexError as err:
violations.append(err)
file_segment = None
else:
file_segment = None
if file_segment:
verbosity_logger(file_segment.stringify(), verbosity=verbosity)
t2 = time.monotonic()
bencher("Lexing {0!r}".format(short_fname))
verbosity_logger("PARSING ({0})".format(fname), verbosity=verbosity)
# Parse the file and log any problems
if file_segment:
try:
# Make a parse context and parse
context = self.get_parse_context()
context.verbosity = verbosity or context.verbosity
context.recurse = recurse or context.recurse
parsed = file_segment.parse(parse_context=context)
except SQLParseError as err:
violations.append(err)
parsed = None
if parsed:
verbosity_logger(frame_msg("Parsed Tree:"), verbosity=verbosity)
verbosity_logger(parsed.stringify(), verbosity=verbosity)
else:
parsed = None
t3 = time.monotonic()
time_dict = {'templating': t1 - t0, 'lexing': t2 - t1, 'parsing': t3 - t2}
bencher("Finish parsing {0!r}".format(short_fname))
return parsed, violations, time_dict
def parse_string(self,
s,
fname=None,
verbosity=0,
recurse=True,
config=None):
"""Parse a string.
Returns:
`tuple` of (`parsed`, `violations`, `time_dict`, `config_diff`).
`parsed` is a segment structure representing the parsed file. If
parsing fails due to an inrecoverable violation then we will
return None.
`violations` is a :obj:`list` of violations so far, which will either be
templating, lexing or parsing violations at this stage.
`time_dict` is a :obj:`dict` containing timings for how long each step
took in the process.
"""
violations = []
t0 = time.monotonic()
bencher = BenchIt() # starts the timer
if fname:
short_fname = fname.replace('\\', '/').split('/')[-1]
else:
# this handles to potential case of a null fname
short_fname = fname
bencher("Staring parse_string for {0!r}".format(short_fname))
# Log the start of this process if we're in a more verbose mode.
if verbosity > 1:
self.log(
format_filename(filename=fname,
success='PARSING',
verbose=verbosity))
# This is where we output config diffs if they exist.
if config:
# Only output config diffs if there is a config to diff to.
config_diff = config.diff_to(self.config)
if config_diff:
self.log(" Config Diff:")
self.log(
format_config_vals(
self.config.iter_vals(cfg=config_diff)))
verbosity_logger("TEMPLATING RAW [{0}] ({1})".format(
self.templater.name, fname),
verbosity=verbosity)
s, templater_violations = self.templater.process(s,
fname=fname,
config=config
or self.config)
violations += templater_violations
# Detect the case of a catastrophic templater fail. In this case
# we don't continue. We'll just bow out now.
if not s:
file_segment = None
t1 = time.monotonic()
bencher("Templating {0!r}".format(short_fname))
if s:
verbosity_logger("LEXING RAW ({0})".format(fname),
verbosity=verbosity)
# Lex the file and log any problems
try:
file_segment, lex_vs = FileSegment.from_raw(s,
config=config
or self.config)
# We might just get the violations as a list
violations += lex_vs
except SQLLexError as err:
violations.append(err)
file_segment = None
else:
file_segment = None
if file_segment:
verbosity_logger(file_segment.stringify(), verbosity=verbosity)
t2 = time.monotonic()
bencher("Lexing {0!r}".format(short_fname))
verbosity_logger("PARSING ({0})".format(fname), verbosity=verbosity)
# Parse the file and log any problems
if file_segment:
try:
# Make a parse context and parse
context = self.get_parse_context(config=config or self.config)
context.verbosity = verbosity or context.verbosity
context.recurse = recurse or context.recurse
parsed = file_segment.parse(parse_context=context)
except SQLParseError as err:
violations.append(err)
parsed = None
if parsed:
verbosity_logger(frame_msg("Parsed Tree:"),
verbosity=verbosity)
verbosity_logger(parsed.stringify(), verbosity=verbosity)
# We may succeed parsing, but still have unparsable segments. Extract them here.
for unparsable in parsed.iter_unparsables():
# No exception has been raised explicitly, but we still create one here
# so that we can use the common interface
violations.append(
SQLParseError("Found unparsable section: {0!r}".format(
unparsable.raw if len(unparsable.raw) < 40 else
unparsable.raw[:40] + "..."),
segment=unparsable))
if verbosity >= 2:
verbosity_logger("Found unparsable segment...",
verbosity=verbosity)
verbosity_logger(unparsable.stringify(),
verbosity=verbosity)
else:
parsed = None
t3 = time.monotonic()
time_dict = {
'templating': t1 - t0,
'lexing': t2 - t1,
'parsing': t3 - t2
}
bencher("Finish parsing {0!r}".format(short_fname))
return parsed, violations, time_dict
def parse_string(self, in_str, fname=None, recurse=True, config=None):
"""Parse a string.
Returns:
`ParsedString` of (`parsed`, `violations`, `time_dict`, `templated_file`).
`parsed` is a segment structure representing the parsed file. If
parsing fails due to an inrecoverable violation then we will
return None.
`violations` is a :obj:`list` of violations so far, which will either be
templating, lexing or parsing violations at this stage.
`time_dict` is a :obj:`dict` containing timings for how long each step
took in the process.
`templated_file` is a :obj:`TemplatedFile` containing the details
of the templated file.
"""
violations = []
t0 = time.monotonic()
bencher = BenchIt() # starts the timer
if fname:
short_fname = fname.replace("\\", "/").split("/")[-1]
else:
# this handles the potential case of a null fname
short_fname = fname
bencher("Staring parse_string for {0!r}".format(short_fname))
# Dispatch the output for the parse header (including the config diff)
if self.formatter:
self.formatter.dispatch_parse_header(fname, self.config, config)
# Just use the local config from here:
config = config or self.config
# Scan the raw file for config commands.
for raw_line in in_str.splitlines():
if raw_line.startswith("-- sqlfluff"):
# Found a in-file config command
config.process_inline_config(raw_line)
linter_logger.info("TEMPLATING RAW [%s] (%s)", self.templater.name,
fname)
templated_file, templater_violations = self.templater.process(
in_str=in_str, fname=fname, config=config)
violations += templater_violations
# Detect the case of a catastrophic templater fail. In this case
# we don't continue. We'll just bow out now.
if not templated_file:
linter_logger.info("TEMPLATING FAILED: %s", templater_violations)
tokens = None
t1 = time.monotonic()
bencher("Templating {0!r}".format(short_fname))
if templated_file:
linter_logger.info("LEXING RAW (%s)", fname)
# Get the lexer
lexer = Lexer(config=config)
# Lex the file and log any problems
try:
tokens, lex_vs = lexer.lex(templated_file)
# We might just get the violations as a list
violations += lex_vs
except SQLLexError as err:
linter_logger.info("LEXING FAILED! (%s): %s", fname, err)
violations.append(err)
tokens = None
else:
tokens = None
if tokens:
linter_logger.info("Lexed tokens: %s", [seg.raw for seg in tokens])
else:
linter_logger.info("NO LEXED TOKENS!")
if tokens:
# Check that we've got sensible indentation from the lexer.
# We might need to supress if it's a complicated file.
templating_blocks_indent = config.get("template_blocks_indent",
"indentation")
if isinstance(templating_blocks_indent, str):
force_block_indent = templating_blocks_indent.lower().strip(
) == "force"
else:
force_block_indent = False
templating_blocks_indent = bool(templating_blocks_indent)
# If we're forcing it through we don't check.
if templating_blocks_indent and not force_block_indent:
indent_balance = sum(
getattr(elem, "indent_val", 0) for elem in tokens)
if indent_balance != 0:
linter_logger.warning(
"Indent balance test failed for %r. Template indents will not be linted for this file.",
fname,
)
# Don't enable the templating blocks.
templating_blocks_indent = False
# Disable the linting of L003 on templated tokens.
config.set_value(
["rules", "L003", "lint_templated_tokens"], False)
# The file will have been lexed without config, so check all indents
# are enabled.
new_tokens = []
for token in tokens:
if token.is_meta:
if token.indent_val != 0:
# Don't allow it if we're not linting templating block indents.
if not templating_blocks_indent:
continue
# Don't allow if it's not configure to function.
elif not token.is_enabled(indent_config=config.
get_section("indentation")):
continue
new_tokens.append(token)
# Swap the buffers
tokens = new_tokens
t2 = time.monotonic()
bencher("Lexing {0!r}".format(short_fname))
linter_logger.info("PARSING (%s)", fname)
parser = Parser(config=config)
# Parse the file and log any problems
if tokens:
try:
parsed = parser.parse(tokens, recurse=recurse)
except SQLParseError as err:
linter_logger.info("PARSING FAILED! (%s): %s", fname, err)
violations.append(err)
parsed = None
if parsed:
linter_logger.info(
"\n###\n#\n# {0}\n#\n###".format("Parsed Tree:"))
linter_logger.info("\n" + parsed.stringify())
# We may succeed parsing, but still have unparsable segments. Extract them here.
for unparsable in parsed.iter_unparsables():
# No exception has been raised explicitly, but we still create one here
# so that we can use the common interface
violations.append(
SQLParseError(
"Found unparsable section: {0!r}".format(
unparsable.raw if len(unparsable.raw) < 40 else
unparsable.raw[:40] + "..."),
segment=unparsable,
))
linter_logger.info("Found unparsable segment...")
linter_logger.info(unparsable.stringify())
else:
parsed = None
t3 = time.monotonic()
time_dict = {
"templating": t1 - t0,
"lexing": t2 - t1,
"parsing": t3 - t2
}
bencher("Finish parsing {0!r}".format(short_fname))
return ParsedString(parsed, violations, time_dict, templated_file,
config)
def parse_string(
self,
in_str: str,
fname: Optional[str] = None,
recurse: bool = True,
config: Optional[FluffConfig] = None,
) -> ParsedString:
"""Parse a string."""
violations: List[SQLBaseError] = []
t0 = time.monotonic()
bencher = BenchIt() # starts the timer
short_fname = self._generate_short_fname(fname)
bencher("Staring parse_string for {0!r}".format(short_fname))
# Dispatch the output for the template header (including the config diff)
if self.formatter:
self.formatter.dispatch_template_header(fname, self.config, config)
# Just use the local config from here:
config = config or self.config
# Scan the raw file for config commands.
config.process_raw_file_for_config(in_str)
templated_file, templater_violations = self.render_string(
in_str, fname, config)
violations += templater_violations
t1 = time.monotonic()
bencher("Templating {0!r}".format(short_fname))
# Dispatch the output for the parse header
if self.formatter:
self.formatter.dispatch_parse_header(fname)
tokens: Optional[Sequence[BaseSegment]]
if templated_file:
tokens, lvs, config = self._lex_templated_file(
templated_file, config)
violations += lvs
else:
tokens = None
t2 = time.monotonic()
bencher("Lexing {0!r}".format(short_fname))
linter_logger.info("PARSING (%s)", fname)
if tokens:
parsed, pvs = self._parse_tokens(tokens, config, recurse=recurse)
violations += pvs
else:
parsed = None
t3 = time.monotonic()
time_dict = {
"templating": t1 - t0,
"lexing": t2 - t1,
"parsing": t3 - t2
}
bencher("Finish parsing {0!r}".format(short_fname))
return ParsedString(parsed, violations, time_dict, templated_file,
config)
def parse_string(self, s, fname=None, recurse=True, config=None):
"""Parse a string.
Returns:
`tuple` of (`parsed`, `violations`, `time_dict`, `config_diff`).
`parsed` is a segment structure representing the parsed file. If
parsing fails due to an inrecoverable violation then we will
return None.
`violations` is a :obj:`list` of violations so far, which will either be
templating, lexing or parsing violations at this stage.
`time_dict` is a :obj:`dict` containing timings for how long each step
took in the process.
"""
violations = []
t0 = time.monotonic()
bencher = BenchIt() # starts the timer
if fname:
short_fname = fname.replace('\\', '/').split('/')[-1]
else:
# this handles to potential case of a null fname
short_fname = fname
bencher("Staring parse_string for {0!r}".format(short_fname))
# Dispatch the output for the parse header (including the config diff)
if self.formatter:
self.formatter.dispatch_parse_header(fname, self.config, config)
linter_logger.info("TEMPLATING RAW [%s] (%s)", self.templater.name,
fname)
s, templater_violations = self.templater.process(s,
fname=fname,
config=config
or self.config)
violations += templater_violations
# Detect the case of a catastrophic templater fail. In this case
# we don't continue. We'll just bow out now.
if not s:
file_segment = None
t1 = time.monotonic()
bencher("Templating {0!r}".format(short_fname))
if s:
linter_logger.info("LEXING RAW (%s)", fname)
# Lex the file and log any problems
try:
file_segment, lex_vs = FileSegment.from_raw(s,
config=config
or self.config)
# We might just get the violations as a list
violations += lex_vs
except SQLLexError as err:
violations.append(err)
file_segment = None
else:
file_segment = None
if file_segment:
linter_logger.info(file_segment.stringify())
t2 = time.monotonic()
bencher("Lexing {0!r}".format(short_fname))
linter_logger.info("PARSING (%s)", fname)
# Parse the file and log any problems
if file_segment:
try:
# Make a parse context and parse
with RootParseContext.from_config(config=config or self.config,
recurse=recurse) as ctx:
parsed = file_segment.parse(parse_context=ctx)
except SQLParseError as err:
violations.append(err)
parsed = None
if parsed:
linter_logger.info(
"\n###\n#\n# {0}\n#\n###".format("Parsed Tree:"))
linter_logger.info("\n" + parsed.stringify())
# We may succeed parsing, but still have unparsable segments. Extract them here.
for unparsable in parsed.iter_unparsables():
# No exception has been raised explicitly, but we still create one here
# so that we can use the common interface
violations.append(
SQLParseError("Found unparsable section: {0!r}".format(
unparsable.raw if len(unparsable.raw) < 40 else
unparsable.raw[:40] + "..."),
segment=unparsable))
linter_logger.info("Found unparsable segment...")
linter_logger.info(unparsable.stringify())
else:
parsed = None
t3 = time.monotonic()
time_dict = {
'templating': t1 - t0,
'lexing': t2 - t1,
'parsing': t3 - t2
}
bencher("Finish parsing {0!r}".format(short_fname))
return parsed, violations, time_dict
def fix(force,
paths,
bench=False,
fixed_suffix="",
no_safety=False,
logger=None,
**kwargs):
"""Fix SQL files.
PATH is the path to a sql file or directory to lint. This can be either a
file ('path/to/file.sql'), a path ('directory/of/sql/files'), a single ('-')
character to indicate reading from *stdin* or a dot/blank ('.'/' ') which will
be interpreted like passing the current working directory as a path argument.
"""
c = get_config(**kwargs)
lnt, formatter = get_linter_and_formatter(c, silent=("-", ) == paths)
verbose = c.get("verbose")
bencher = BenchIt()
formatter.dispatch_config(lnt)
# Set up logging.
set_logging_level(verbosity=verbose, logger=logger)
# Check that if fix is specified, that we have picked only a subset of rules
if no_safety:
click.echo(
colorize("NO SAFETY", "red") +
": Attempting fixes for all enabled rules.")
elif lnt.config.get("rule_whitelist") is None:
click.echo(("The fix option is only available in combination"
" with --rules. This is for your own safety! To"
" disable this safety feature use --no-safety or --s."))
sys.exit(1)
# handle stdin case. should output formatted sql to stdout and nothing else.
if ("-", ) == paths:
stdin = sys.stdin.read()
# TODO: Remove verbose
result = lnt.lint_string_wrapped(stdin, fname="stdin", fix=True)
stdout = result.paths[0].files[0].fix_string()[0]
click.echo(stdout, nl=False)
sys.exit()
# Lint the paths (not with the fix argument at this stage), outputting as we go.
click.echo("==== finding fixable violations ====")
try:
result = lnt.lint_paths(paths,
fix=True,
ignore_non_existent_files=False)
except IOError:
click.echo(
colorize(
"The path(s) {0!r} could not be accessed. Check it/they exist(s)."
.format(paths),
"red",
))
sys.exit(1)
# NB: We filter to linting violations here, because they're
# the only ones which can be potentially fixed.
if result.num_violations(types=SQLLintError, fixable=True) > 0:
click.echo("==== fixing violations ====")
click.echo("{0} fixable linting violations found".format(
result.num_violations(types=SQLLintError, fixable=True)))
if force:
click.echo(colorize("FORCE MODE", "red") + ": Attempting fixes...")
# TODO: Remove verbose
success = do_fixes(
lnt,
result,
formatter,
types=SQLLintError,
fixed_file_suffix=fixed_suffix,
)
if not success:
sys.exit(1)
else:
click.echo("Are you sure you wish to attempt to fix these? [Y/n] ",
nl=False)
c = click.getchar().lower()
click.echo("...")
if c == "y":
click.echo("Attempting fixes...")
# TODO: Remove verbose
success = do_fixes(
lnt,
result,
formatter,
types=SQLLintError,
fixed_file_suffix=fixed_suffix,
)
if not success:
sys.exit(1)
elif c == "n":
click.echo("Aborting...")
else:
click.echo("Invalid input :(")
click.echo("Aborting...")
else:
click.echo("==== no fixable linting violations found ====")
if result.num_violations(types=SQLLintError, fixable=False) > 0:
click.echo(" [{0} unfixable linting violations found]".format(
result.num_violations(types=SQLLintError, fixable=False)))
if bench:
click.echo("\n\n==== bencher stats ====")
bencher.display()
sys.exit(0)
def fix_string(self, verbosity=0):
"""Obtain the changes to a path as a string.
We use the file_mask to do a safe merge, avoiding any templated
sections. First we need to detect where there have been changes
between the fixed and templated versions. The file mask is of
the format: (raw_file, templated_file, fixed_file).
We use difflib.SequenceMatcher.get_opcodes
See: https://docs.python.org/3.7/library/difflib.html#difflib.SequenceMatcher.get_opcodes
It returns a list of tuples ('equal|replace|delete|insert', ia1, ia2, ib1, ib2).
"""
bencher = BenchIt()
bencher("fix_string: start")
# Do we have enough information to actually fix the file?
if any(elem is None for elem in self.file_mask):
verbosity_logger(
"Insufficient information to fix file: {0}".format(
self.file_mask),
verbosity=verbosity)
return None, False
verbosity_logger("Persisting file masks: {0}".format(self.file_mask),
verbosity=verbosity)
# Compare Templated with Raw
diff_templ = SequenceMatcher(autojunk=None,
a=self.file_mask[0],
b=self.file_mask[1])
bencher("fix_string: Match 0&1")
diff_templ_codes = diff_templ.get_opcodes()
verbosity_logger("Templater diff codes: {0}".format(diff_templ_codes),
verbosity=verbosity)
bencher("fix_string: Got Opcodes 0&1")
# Compare Fixed with Templated
diff_fix = SequenceMatcher(autojunk=None,
a=self.file_mask[1],
b=self.file_mask[2])
bencher("fix_string: Matched 1&2")
# diff_fix = SequenceMatcher(autojunk=None, a=self.file_mask[1][0], b=self.file_mask[2][0])
diff_fix_codes = diff_fix.get_opcodes()
verbosity_logger("Fixing diff codes: {0}".format(diff_fix_codes),
verbosity=verbosity)
bencher("fix_string: Got Opcodes 1&2")
# If diff_templ isn't the same then we should just keep the template. If there *was*
# a fix in that space, then we should raise an issue
# If it is the same, then we can apply fixes as expected.
write_buff = ''
fixed_block = None
templ_block = None
# index in raw, templ and fix
idx = (0, 0, 0)
loop_idx = 0
bencher("fix_string: Loop Setup")
while True:
loop_idx += 1
verbosity_logger("{0:04d}: Write Loop: idx:{1}, buff:{2!r}".format(
loop_idx, idx, write_buff),
verbosity=verbosity)
if templ_block is None:
if diff_templ_codes:
templ_block = diff_templ_codes.pop(0)
# We've exhausted the template. Have we exhausted the fixes?
elif fixed_block is None and not diff_fix_codes:
# Yes - excellent. DONE
break
# Deal with the case that we only have inserts left.
elif all(elem[0] == 'insert' for elem in diff_fix_codes):
for fixed_block in diff_fix_codes:
write_buff += self.file_mask[2][
fixed_block[3]:fixed_block[4]]
break
else:
raise NotImplementedError(
"Fix Block(s) left over! Don't know how to handle this! aeflf8wh"
)
if fixed_block is None:
if diff_fix_codes:
fixed_block = diff_fix_codes.pop(0)
# One case is that we just consumed the last block of both, so check indexes
# to see if we're at the end of the raw file.
elif idx[0] >= len(self.file_mask[0]):
# Yep we're at the end
break
else:
raise NotImplementedError(
"Unexpectedly depleted the fixes. Panic!")
verbosity_logger("{0:04d}: Blocks: template:{1}, fix:{2}".format(
loop_idx, templ_block, fixed_block),
verbosity=verbosity)
if templ_block[0] == 'equal':
if fixed_block[0] == 'equal':
# No templating, no fixes, go with middle and advance indexes
# Find out how far we can advance (we use the middle version because it's common)
if templ_block[4] == fixed_block[2]:
buff = self.file_mask[1][idx[1]:fixed_block[2]]
# consume both blocks
fixed_block = None
templ_block = None
elif templ_block[4] > fixed_block[2]:
buff = self.file_mask[1][idx[1]:fixed_block[2]]
# consume fixed block
fixed_block = None
elif templ_block[4] < fixed_block[2]:
buff = self.file_mask[1][idx[1]:templ_block[4]]
# consume templ block
templ_block = None
idx = (idx[0] + len(buff), idx[1] + len(buff),
idx[2] + len(buff))
write_buff += buff
continue
elif fixed_block[0] == 'replace':
# Consider how to apply fixes.
# Can we implement the fix while staying in the equal segment?
if fixed_block[2] <= templ_block[4]:
# Yes! Write from the fixed version.
write_buff += self.file_mask[2][idx[2]:fixed_block[4]]
idx = (idx[0] + (fixed_block[2] - fixed_block[1]),
fixed_block[2], fixed_block[4])
# Consume the fixed block because we've written the whole thing.
fixed_block = None
continue
else:
raise NotImplementedError("DEF")
elif fixed_block[0] == 'delete':
# We're deleting items, nothing to write but we can consume some
# blocks and advance some indexes.
idx = (idx[0] + (fixed_block[2] - fixed_block[1]),
fixed_block[2], fixed_block[4])
fixed_block = None
elif fixed_block[0] == 'insert':
# We're inserting items, Write from the fix block, but only that index moves.
write_buff += self.file_mask[2][idx[2]:fixed_block[4]]
idx = (idx[0], idx[1], fixed_block[4])
fixed_block = None
else:
raise NotImplementedError((
"Unexpected opcode {0} for fix block! Please report this "
"issue on github with the query and rules you're trying to "
"fix.").format(fixed_block[0]))
elif templ_block[0] == 'replace':
# We're in a templated section - we should write the templated version.
# we should consume the whole replace block and then deal with where
# we end up.
buff = self.file_mask[0][idx[0]:templ_block[2]]
new_templ_idx = templ_block[4]
# Fast forward through fix blocks until we catch up. We're not implementing
# any changes in a templated section.
while True:
if fixed_block[2] > new_templ_idx >= fixed_block[1]:
# this block contains the end point
break
else:
# We're not at the end point yet, continue to fast forward through.
if fixed_block[0] != 'equal':
print("WARNING: Skipping edit block: {0}".format(
fixed_block))
if diff_fix_codes:
fixed_block = diff_fix_codes.pop(0)
else:
raise NotImplementedError(
"Unexpectedly depleted the fixes. Panic!")
# Are we exactly on a join?
if new_templ_idx == fixed_block[1]:
# GREAT - this makes things easy because we have an equality point already
idx = (templ_block[2], new_templ_idx, fixed_block[3])
else:
if fixed_block[0] == 'equal':
# If it's in an equal block, we can use the same offset from the end.
idx = (templ_block[2], new_templ_idx, fixed_block[3] +
(new_templ_idx - fixed_block[1]))
else:
# TODO: We're trying to move through an templated section, but end up
# in a fixed section. We've lost track of indexes.
# We might need to panic if this happens...
print("UMMMMMM!")
print(new_templ_idx)
print(fixed_block)
raise NotImplementedError("ABC")
write_buff += buff
# consume template block
templ_block = None
elif templ_block[0] == 'delete':
# The comparison, things that the templater has deleted
# some characters. This is just a quirk of the differ.
# In reality this means we just write these characters
# and don't worry about advancing the other indexes.
buff = self.file_mask[0][idx[0]:templ_block[2]]
# consume templ block
templ_block = None
idx = (idx[0] + len(buff), idx[1], idx[2])
write_buff += buff
elif templ_block[0] == 'insert':
# The templater has inserted something here. We don't need
# to write anything here (because whatever we're looking at
# was inserted by the templater), but we do need to keep
# track of what happened to the rest of the section we're in.
# If nothing was fixed then it's easy because the indices
# will be the same. Otherwise... great question...
# For now let's just deal with the happy case where the fixed
# block is equal
if fixed_block[0] == 'equal':
# Let's make sure we can consume enough to get through the
# templ block and not get to the end of the fix block.
if templ_block[4] <= fixed_block[2]:
insert_len = templ_block[4] - templ_block[3]
idx = (idx[0], idx[1] + insert_len,
idx[2] + insert_len)
# if things matched up perfectly, consume the fixed block
if templ_block[4] == fixed_block[2]:
fixed_block = None
# always consume templ block in this case
templ_block = None
else:
raise NotImplementedError((
"Unexpected scenario during insert opcode! Please report "
"this issue on github with the query and rules you're trying "
"to fix."))
else:
raise NotImplementedError((
"Unexpected opcode {0} for fix block! Please report this "
"issue on github with the query and rules you're trying to "
"fix.").format(fixed_block[0]))
else:
raise NotImplementedError((
"Unexpected opcode {0} for template block! Please report this "
"issue on github with the query and rules you're trying to "
"fix.").format(templ_block[0]))
bencher("fix_string: Fixing loop done")
# The success metric here is whether anything ACTUALLY changed.
return write_buff, write_buff != self.file_mask[0]
def parse(path,
code_only,
format,
profiler,
bench,
nofail,
logger=None,
**kwargs):
"""Parse SQL files and just spit out the result.
PATH is the path to a sql file or directory to lint. This can be either a
file ('path/to/file.sql'), a path ('directory/of/sql/files'), a single ('-')
character to indicate reading from *stdin* or a dot/blank ('.'/' ') which will
be interpreted like passing the current working directory as a path argument.
"""
# Initialise the benchmarker
bencher = BenchIt() # starts the timer
c = get_config(**kwargs)
# We don't want anything else to be logged if we want a yaml output
lnt, formatter = get_linter_and_formatter(c,
silent=format
in ("json", "yaml"))
verbose = c.get("verbose")
recurse = c.get("recurse")
formatter.dispatch_config(lnt)
# Set up logging.
set_logging_level(verbosity=verbose, logger=logger)
# TODO: do this better
nv = 0
if profiler:
# Set up the profiler if required
try:
import cProfile
except ImportError:
click.echo("The cProfiler is not available on your platform.")
sys.exit(1)
pr = cProfile.Profile()
pr.enable()
bencher("Parse setup")
try:
# handle stdin if specified via lone '-'
if "-" == path:
# put the parser result in a list to iterate later
config = lnt.config.make_child_from_path("stdin")
result = [(
# TODO: Remove verbose
*lnt.parse_string(
sys.stdin.read(), "stdin", recurse=recurse, config=config),
config,
)]
else:
# A single path must be specified for this command
# TODO: Remove verbose
result = lnt.parse_path(path, recurse=recurse)
# iterative print for human readout
if format == "human":
for parsed, violations, time_dict, f_cfg in result:
if parsed:
click.echo(parsed.stringify(code_only=code_only))
else:
# TODO: Make this prettier
click.echo("...Failed to Parse...")
nv += len(violations)
if violations:
click.echo("==== parsing violations ====")
for v in violations:
click.echo(format_violation(v))
if violations and f_cfg.get("dialect") == "ansi":
click.echo(format_dialect_warning())
if verbose >= 2:
click.echo("==== timings ====")
click.echo(cli_table(time_dict.items()))
bencher("Output details for file")
else:
# collect result and print as single payload
# will need to zip in the file paths
filepaths = ["stdin"] if "-" == path else lnt.paths_from_path(path)
result = [
dict(
filepath=filepath,
segments=parsed.as_record(code_only=code_only,
show_raw=True),
) for filepath, (parsed, _, _, _) in zip(filepaths, result)
]
if format == "yaml":
# For yaml dumping always dump double quoted strings if they contain tabs or newlines.
def quoted_presenter(dumper, data):
"""Representer which always double quotes string values needing escapes."""
if "\n" in data or "\t" in data or "'" in data:
return dumper.represent_scalar("tag:yaml.org,2002:str",
data,
style='"')
else:
return dumper.represent_scalar("tag:yaml.org,2002:str",
data,
style="")
yaml.add_representer(str, quoted_presenter)
click.echo(yaml.dump(result))
elif format == "json":
click.echo(json.dumps(result))
except IOError:
click.echo(
colorize(
"The path {0!r} could not be accessed. Check it exists.".
format(path),
"red",
))
sys.exit(1)
if profiler:
pr.disable()
profiler_buffer = StringIO()
ps = pstats.Stats(pr, stream=profiler_buffer).sort_stats("cumulative")
ps.print_stats()
click.echo("==== profiler stats ====")
# Only print the first 50 lines of it
click.echo("\n".join(profiler_buffer.getvalue().split("\n")[:50]))
if bench:
click.echo("\n\n==== bencher stats ====")
bencher.display()
if nv > 0 and not nofail:
sys.exit(66)
else:
sys.exit(0)
def fix_string(self):
"""Obtain the changes to a path as a string.
We use the source mapping features of TemplatedFile
to generate a list of "patches" which cover the non
templated parts of the file and refer back to the locations
in the original file.
NB: This is MUCH FASTER than the original approach
using difflib in pre 0.4.0.
There is an important distinction here between Slices and
Segments. A Slice is a portion of a file which is determined
by the templater based on which portions of the source file
are templated or not, and therefore before Lexing and so is
completely dialect agnostic. A Segment is determined by the
Lexer from portions of strings after templating.
"""
bencher = BenchIt()
bencher("fix_string: start")
linter_logger.debug("Original Tree: %r",
self.templated_file.templated_str)
linter_logger.debug("Fixed Tree: %r", self.tree.raw)
# The sliced file is contigious in the TEMPLATED space.
# NB: It has gaps and repeats in the source space.
# It's also not the FIXED file either.
linter_logger.debug("### Templated File.")
for idx, file_slice in enumerate(self.templated_file.sliced_file):
t_str = self.templated_file.templated_str[
file_slice.templated_slice]
s_str = self.templated_file.source_str[file_slice.source_slice]
if t_str == s_str:
linter_logger.debug(" File slice: %s %r [invariant]", idx,
file_slice)
else:
linter_logger.debug(" File slice: %s %r", idx, file_slice)
linter_logger.debug(" \t\t\ttemplated: %r\tsource: %r",
t_str, s_str)
original_source = self.templated_file.source_str
# Make sure no patches overlap and divide up the source file into slices.
# Any Template tags in the source file are off limits.
source_only_slices = self.templated_file.source_only_slices()
linter_logger.debug("Source-only slices: %s", source_only_slices)
# Iterate patches, filtering and translating as we go:
linter_logger.debug("### Beginning Patch Iteration.")
filtered_source_patches = []
dedupe_buffer = []
# We use enumerate so that we get an index for each patch. This is entirely
# so when debugging logs we can find a given patch again!
for idx, patch in enumerate(
self.tree.iter_patches(
templated_str=self.templated_file.templated_str)):
linter_logger.debug(" %s Yielded patch: %s", idx, patch)
# This next bit is ALL FOR LOGGING AND DEBUGGING
if patch.templated_slice.start >= 10:
pre_hint = self.templated_file.templated_str[
patch.templated_slice.start -
10:patch.templated_slice.start]
else:
pre_hint = self.templated_file.templated_str[:patch.
templated_slice.
start]
if patch.templated_slice.stop + 10 < len(
self.templated_file.templated_str):
post_hint = self.templated_file.templated_str[
patch.templated_slice.stop:patch.templated_slice.stop + 10]
else:
post_hint = self.templated_file.templated_str[
patch.templated_slice.stop:]
linter_logger.debug(" Templated Hint: ...%r <> %r...",
pre_hint, post_hint)
# Attempt to convert to source space.
try:
source_slice = self.templated_file.templated_slice_to_source_slice(
patch.templated_slice, )
except ValueError:
linter_logger.info(
" - Skipping. Source space Value Error. i.e. attempted insertion within templated section."
)
# If we try and slice within a templated section, then we may fail
# in which case, we should skip this patch.
continue
# Check for duplicates
dedupe_tuple = (source_slice, patch.fixed_raw)
if dedupe_tuple in dedupe_buffer:
linter_logger.info(
" - Skipping. Source space Duplicate: %s",
dedupe_tuple)
continue
# We now evaluate patches in the source-space for whether they overlap
# or disrupt any templated sections.
# The intent here is that unless explicity stated, a fix should never
# disrupt a templated section.
# NOTE: We rely here on the patches being sorted.
# TODO: Implement a mechanism for doing templated section fixes. For
# now it's just not allowed.
# Get the affected raw slices.
local_raw_slices = self.templated_file.raw_slices_spanning_source_slice(
source_slice)
local_type_list = [slc.slice_type for slc in local_raw_slices]
enriched_patch = EnrichedFixPatch(
source_slice=source_slice,
templated_slice=patch.templated_slice,
patch_type=patch.patch_type,
fixed_raw=patch.fixed_raw,
templated_str=self.templated_file.templated_str[
patch.templated_slice],
source_str=self.templated_file.source_str[source_slice],
)
# Deal with the easy case of only literals
if set(local_type_list) == {"literal"}:
linter_logger.info(
" * Keeping patch on literal-only section: %s",
enriched_patch)
filtered_source_patches.append(enriched_patch)
dedupe_buffer.append(enriched_patch.dedupe_tuple())
# Is it a zero length pathch.
elif (enriched_patch.source_slice.start
== enriched_patch.source_slice.stop
and enriched_patch.source_slice.start
== local_raw_slices[0].source_idx):
linter_logger.info(
" * Keeping insertion patch on slice boundary: %s",
enriched_patch,
)
filtered_source_patches.append(enriched_patch)
dedupe_buffer.append(enriched_patch.dedupe_tuple())
# If it's ONLY templated then we should skip it.
elif "literal" not in local_type_list:
linter_logger.info(
" - Skipping patch over templated section: %s",
enriched_patch)
# If we span more than two slices then we should just skip it. Too Hard.
elif len(local_raw_slices) > 2:
linter_logger.info(
" - Skipping patch over more than two raw slices: %s",
enriched_patch,
)
# If it's an insertion (i.e. the string in the pre-fix template is '') then we
# won't be able to place it, so skip.
elif not enriched_patch.templated_str:
linter_logger.info(
" - Skipping insertion patch in templated section: %s",
enriched_patch,
)
# If the string from the templated version isn't in the source, then we can't fix it.
elif enriched_patch.templated_str not in enriched_patch.source_str:
linter_logger.info(
" - Skipping edit patch on templated content: %s",
enriched_patch,
)
else:
# Identify all the places the string appears in the source content.
positions = list(
findall(enriched_patch.templated_str,
enriched_patch.source_str))
if len(positions) != 1:
linter_logger.debug(
" - Skipping edit patch on non-unique templated content: %s",
enriched_patch,
)
continue
# We have a single occurances of the thing we want to patch. This
# means we can use it's position to place our patch.
new_source_slice = slice(
enriched_patch.source_slice.start + positions[0],
enriched_patch.source_slice.start + positions[0] +
len(enriched_patch.templated_str),
)
enriched_patch = EnrichedFixPatch(
source_slice=new_source_slice,
templated_slice=enriched_patch.templated_slice,
patch_type=enriched_patch.patch_type,
fixed_raw=enriched_patch.fixed_raw,
templated_str=enriched_patch.templated_str,
source_str=enriched_patch.source_str,
)
linter_logger.debug(
" * Keeping Tricky Case. Positions: %s, New Slice: %s, Patch: %s",
positions,
new_source_slice,
enriched_patch,
)
filtered_source_patches.append(enriched_patch)
dedupe_buffer.append(enriched_patch.dedupe_tuple())
continue
# Sort the patches before building up the file.
filtered_source_patches = sorted(filtered_source_patches,
key=lambda x: x.source_slice.start)
# We now slice up the file using the patches and any source only slices.
# This gives us regions to apply changes to.
slice_buff = []
source_idx = 0
for patch in filtered_source_patches:
# Are there templated slices at or before the start of this patch?
while (source_only_slices and source_only_slices[0].source_idx <
patch.source_slice.start):
next_so_slice = source_only_slices.pop(0).source_slice()
# Add a pre-slice before the next templated slices if needed.
if next_so_slice.start > source_idx:
slice_buff.append(slice(source_idx, next_so_slice.start))
# Add the templated slice.
slice_buff.append(next_so_slice)
source_idx = next_so_slice.stop
# Is there a gap between current position and this patch?
if patch.source_slice.start > source_idx:
# Add a slice up to this patch.
slice_buff.append(slice(source_idx, patch.source_slice.start))
# Is this patch covering an area we've already covered?
if patch.source_slice.start < source_idx:
linter_logger.info(
"Skipping overlapping patch at Index %s, Patch: %s",
source_idx,
patch,
)
# Ignore the patch for now...
continue
# Add this patch.
slice_buff.append(patch.source_slice)
source_idx = patch.source_slice.stop
# Add a tail slice.
if source_idx < len(self.templated_file.source_str):
slice_buff.append(
slice(source_idx, len(self.templated_file.source_str)))
linter_logger.debug("Final slice buffer: %s", slice_buff)
# Iterate through the patches, building up the new string.
str_buff = ""
for source_slice in slice_buff:
# Is it one in the patch buffer:
for patch in filtered_source_patches:
if patch.source_slice == source_slice:
# Use the patched version
linter_logger.debug(
"%-30s %s %r > %r",
"Appending {} Patch:".format(patch.patch_type),
patch.source_slice,
patch.source_str,
patch.fixed_raw,
)
str_buff += patch.fixed_raw
break
else:
# Use the raw string
linter_logger.debug(
"Appending Raw: %s %r",
source_slice,
self.templated_file.source_str[source_slice],
)
str_buff += self.templated_file.source_str[source_slice]
bencher("fix_string: Fixing loop done")
# The success metric here is whether anything ACTUALLY changed.
return str_buff, str_buff != original_source
