def _get_segment_move_context(
self, target_segment: RawSegment,
parent_segment: BaseSegment) -> SegmentMoveContext:
# Locate the segment to be moved (i.e. context.segment) and search back
# over the raw stack to find the end of the preceding statement.
reversed_raw_stack = Segments(*parent_segment.raw_segments).reversed()
before_code = reversed_raw_stack.select(loop_while=sp.not_(
sp.is_code()),
start_seg=target_segment)
before_segment = before_code.select(sp.not_(sp.is_meta()))
# We're selecting from the raw stack, so we know that before_code is
# made of RawSegment elements.
anchor_segment = (cast(RawSegment, before_code[-1])
if before_code else target_segment)
first_code = reversed_raw_stack.select(
sp.is_code(), start_seg=target_segment).first()
self.logger.debug("Semicolon: first_code: %s", first_code)
is_one_line = (self._is_one_line_statement(
parent_segment, first_code[0]) if first_code else False)
# We can tidy up any whitespace between the segment
# and the preceding code/comment segment.
# Don't mess with comment spacing/placement.
whitespace_deletions = before_segment.select(
loop_while=sp.is_whitespace())
return SegmentMoveContext(anchor_segment, is_one_line, before_segment,
whitespace_deletions)
def _eval(self, context: RuleContext) -> Optional[LintResult]:
"""Find rule violations and provide fixes.
0. Look for a case expression
1. Look for "ELSE"
2. Mark "ELSE" for deletion (populate "fixes")
3. Backtrack and mark all newlines/whitespaces for deletion
4. Look for a raw "NULL" segment
5.a. The raw "NULL" segment is found, we mark it for deletion and return
5.b. We reach the end of case when without matching "NULL": the rule passes
"""
assert context.segment.is_type("case_expression")
children = FunctionalContext(context).segment.children()
else_clause = children.first(sp.is_type("else_clause"))
# Does the "ELSE" have a "NULL"? NOTE: Here, it's safe to look for
# "NULL", as an expression would *contain* NULL but not be == NULL.
if else_clause and else_clause.children(
lambda child: child.raw_upper == "NULL"):
# Found ELSE with NULL. Delete the whole else clause as well as
# indents/whitespaces/meta preceding the ELSE. :TRICKY: Note
# the use of reversed() to make select() effectively search in
# reverse.
before_else = children.reversed().select(
start_seg=else_clause[0],
loop_while=sp.or_(sp.is_name("whitespace", "newline"),
sp.is_meta()),
)
return LintResult(
anchor=context.segment,
fixes=[LintFix.delete(else_clause[0])] +
[LintFix.delete(seg) for seg in before_else],
)
return None
def _get_subsequent_whitespace(
self,
context,
) -> Tuple[Optional[BaseSegment], Optional[BaseSegment]]:
"""Search forwards through the raw segments for subsequent whitespace.
Return a tuple of both the trailing whitespace segment and the
first non-whitespace segment discovered.
"""
# Get all raw segments. "raw_segments" is appropriate as the
# only segments we can care about are comma, whitespace,
# newline, and comment, which are all raw. Using the
# raw_segments allows us to account for possible unexpected
# parse tree structures resulting from other rule fixes.
raw_segments = FunctionalContext(context).raw_segments
# Start after the current comma within the list. Get all the
# following whitespace.
following_segments = raw_segments.select(
loop_while=sp.or_(sp.is_meta(), sp.is_type("whitespace")),
start_seg=context.segment,
)
subsequent_whitespace = following_segments.last(
sp.is_type("whitespace"))
try:
return (
subsequent_whitespace[0] if subsequent_whitespace else None,
raw_segments[raw_segments.index(context.segment) +
len(following_segments) + 1],
)
except IndexError:
# If we find ourselves here it's all whitespace (or nothing) to the
# end of the file. This can only happen in bigquery (see
# test_pass_bigquery_trailing_comma).
return subsequent_whitespace, None
def _get_indexes(context: RuleContext):
children = FunctionalContext(context).segment.children()
select_targets = children.select(sp.is_type("select_clause_element"))
first_select_target_idx = children.find(select_targets.get())
selects = children.select(sp.is_keyword("select"))
select_idx = children.find(selects.get()) if selects else -1
newlines = children.select(sp.is_type("newline"))
first_new_line_idx = children.find(newlines.get()) if newlines else -1
comment_after_select_idx = -1
if newlines:
comment_after_select = children.select(
sp.is_type("comment"),
start_seg=selects.get(),
stop_seg=newlines.get(),
loop_while=sp.or_(
sp.is_type("comment"), sp.is_type("whitespace"), sp.is_meta()
),
)
if comment_after_select:
comment_after_select_idx = (
children.find(comment_after_select.get())
if comment_after_select
else -1
)
first_whitespace_idx = -1
if first_new_line_idx != -1:
# TRICKY: Ignore whitespace prior to the first newline, e.g. if
# the line with "SELECT" (before any select targets) has trailing
# whitespace.
segments_after_first_line = children.select(
sp.is_type("whitespace"), start_seg=children[first_new_line_idx]
)
first_whitespace_idx = children.find(segments_after_first_line.get())
siblings_post = FunctionalContext(context).siblings_post
from_segment = siblings_post.first(sp.is_type("from_clause")).first().get()
pre_from_whitespace = siblings_post.select(
sp.is_type("whitespace"), stop_seg=from_segment
)
return SelectTargetsInfo(
select_idx,
first_new_line_idx,
first_select_target_idx,
first_whitespace_idx,
comment_after_select_idx,
select_targets,
from_segment,
list(pre_from_whitespace),
)
def _report_unused_alias(cls, alias: AliasInfo) -> LintResult:
fixes = [LintFix.delete(alias.alias_expression)] # type: ignore
# Walk back to remove indents/whitespaces
to_delete = (
Segments(
*alias.from_expression_element.segments).reversed().select(
start_seg=alias.alias_expression,
# Stop once we reach an other, "regular" segment.
loop_while=sp.or_(sp.is_whitespace(), sp.is_meta()),
))
fixes += [LintFix.delete(seg) for seg in to_delete]
return LintResult(
anchor=alias.segment,
description="Alias {!r} is never used in SELECT statement.".format(
alias.ref_str),
fixes=fixes,
)
def _eval(self, context: RuleContext) -> Optional[LintResult]:
"""Files must not begin with newlines or whitespace."""
# Only check raw segments. This ensures we don't try and delete the same
# whitespace multiple times (i.e. for non-raw segments higher in the
# tree).
raw_segments = []
whitespace_types = {"newline", "whitespace", "indent", "dedent"}
for seg in context.segment.recursive_crawl_all():
if not seg.is_raw():
continue
if seg.is_type(*whitespace_types):
raw_segments.append(seg)
continue
segment = Segments(seg)
raw_stack = Segments(*raw_segments,
templated_file=context.templated_file)
# Non-whitespace segment.
if (not raw_stack.all(sp.is_meta())
# Found leaf of parse tree.
and not segment.all(sp.is_expandable())
# It is possible that a template segment (e.g.
# {{ config(materialized='view') }}) renders to an empty string
# and as such is omitted from the parsed tree. We therefore
# should flag if a templated raw slice intersects with the
# source slices in the raw stack and skip this rule to avoid
# risking collisions with template objects.
and not raw_stack.raw_slices.any(
rsp.is_slice_type("templated"))):
return LintResult(
anchor=context.segment,
fixes=[LintFix.delete(d) for d in raw_stack],
)
else:
break
return None
def _eval(self, context: RuleContext) -> Optional[LintResult]:
"""Find rule violations and provide fixes."""
# Config type hints
self.prefer_count_0: bool
self.prefer_count_1: bool
if (
# We already know we're in a function because of the crawl_behaviour
context.segment.get_child("function_name").raw_upper == "COUNT"
):
# Get bracketed content
f_content = (FunctionalContext(context).segment.children(
sp.is_type("bracketed")).children(
sp.and_(
sp.not_(sp.is_meta()),
sp.not_(
sp.is_type("start_bracket", "end_bracket",
"whitespace", "newline")),
)))
if len(f_content) != 1: # pragma: no cover
return None
preferred = "*"
if self.prefer_count_1:
preferred = "1"
elif self.prefer_count_0:
preferred = "0"
if f_content[0].is_type("star") and (self.prefer_count_1
or self.prefer_count_0):
return LintResult(
anchor=context.segment,
fixes=[
LintFix.replace(
f_content[0],
[
f_content[0].edit(f_content[0].raw.replace(
"*", preferred))
],
),
],
)
if f_content[0].is_type("expression"):
expression_content = [
seg for seg in f_content[0].segments if not seg.is_meta
]
if (len(expression_content) == 1
and expression_content[0].is_type("literal")
and expression_content[0].raw in ["0", "1"]
and expression_content[0].raw != preferred):
return LintResult(
anchor=context.segment,
fixes=[
LintFix.replace(
expression_content[0],
[
expression_content[0].edit(
expression_content[0].raw.replace(
expression_content[0].raw,
preferred)),
],
),
],
)
return None
def _eval(self, context: RuleContext) -> Optional[LintResult]:
"""Select clause modifiers must appear on same line as SELECT."""
# We only care about select_clause.
assert context.segment.is_type("select_clause")
# Get children of select_clause and the corresponding select keyword.
child_segments = FunctionalContext(context).segment.children()
select_keyword = child_segments[0]
# See if we have a select_clause_modifier.
select_clause_modifier_seg = child_segments.first(
sp.is_type("select_clause_modifier"))
# Rule doesn't apply if there's no select clause modifier.
if not select_clause_modifier_seg:
return None
select_clause_modifier = select_clause_modifier_seg[0]
# Are there any newlines between the select keyword
# and the select clause modifier.
leading_newline_segments = child_segments.select(
select_if=sp.is_type("newline"),
loop_while=sp.or_(sp.is_whitespace(), sp.is_meta()),
start_seg=select_keyword,
)
# Rule doesn't apply if select clause modifier
# is already on the same line as the select keyword.
if not leading_newline_segments:
return None
# We should check if there is whitespace before the select clause modifier
# and remove this during the lint fix.
leading_whitespace_segments = child_segments.select(
select_if=sp.is_type("whitespace"),
loop_while=sp.or_(sp.is_whitespace(), sp.is_meta()),
start_seg=select_keyword,
)
# We should also check if the following select clause element
# is on the same line as the select clause modifier.
trailing_newline_segments = child_segments.select(
select_if=sp.is_type("newline"),
loop_while=sp.or_(sp.is_whitespace(), sp.is_meta()),
start_seg=select_clause_modifier,
)
# We will insert these segments directly after the select keyword.
edit_segments = [
WhitespaceSegment(),
select_clause_modifier,
]
if not trailing_newline_segments:
# if the first select clause element is on the same line
# as the select clause modifier then also insert a newline.
edit_segments.append(NewlineSegment())
fixes = []
# Move select clause modifier after select keyword.
fixes.append(
LintFix.create_after(
anchor_segment=select_keyword,
edit_segments=edit_segments,
))
# Delete original newlines and whitespace between select keyword
# and select clause modifier.
# If there is not a newline after the select clause modifier then delete
# newlines between the select keyword and select clause modifier.
if not trailing_newline_segments:
fixes.extend(LintFix.delete(s) for s in leading_newline_segments)
# If there is a newline after the select clause modifier then delete both the
# newlines and whitespace between the select keyword and select clause modifier.
else:
fixes.extend(
LintFix.delete(s) for s in leading_newline_segments +
leading_whitespace_segments)
# Delete the original select clause modifier.
fixes.append(LintFix.delete(select_clause_modifier))
# If there is whitespace (on the same line) after the select clause modifier
# then also delete this.
trailing_whitespace_segments = child_segments.select(
select_if=sp.is_whitespace(),
loop_while=sp.or_(sp.is_type("whitespace"), sp.is_meta()),
start_seg=select_clause_modifier,
)
if trailing_whitespace_segments:
fixes.extend(
(LintFix.delete(s) for s in trailing_whitespace_segments))
return LintResult(
anchor=context.segment,
fixes=fixes,
)