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 """ if context.segment.is_type("case_expression"): children = context.functional.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 _handle_semicolon_newline( self, context: RuleContext, info: SegmentMoveContext) -> Optional[LintResult]: # Adjust before_segment and anchor_segment for preceding inline # comments. Inline comments can contain noqa logic so we need to add the # newline after the inline comment. ( before_segment, anchor_segment, ) = self._handle_preceding_inline_comments(info.before_segment, info.anchor_segment) if (len(before_segment) == 1) and all( s.is_type("newline") for s in before_segment): return None # If preceding segment is not a single newline then delete the old # semi-colon/preceding whitespace and then insert the # semi-colon in the correct location. # This handles an edge case in which an inline comment comes after # the semi-colon. anchor_segment = self._handle_trailing_inline_comments( context, anchor_segment) fixes = [] if anchor_segment is context.segment: fixes.append( LintFix.replace( anchor_segment, [ NewlineSegment(), SymbolSegment(raw=";", type="symbol", name="semicolon"), ], )) else: fixes.extend([ LintFix.replace( anchor_segment, [ anchor_segment, NewlineSegment(), SymbolSegment(raw=";", type="symbol", name="semicolon"), ], ), LintFix.delete(context.segment, ), ]) fixes.extend(LintFix.delete(d) for d in info.whitespace_deletions) return LintResult( anchor=anchor_segment, fixes=fixes, )
def _eval_multiple_select_target_elements(self, select_targets_info, segment): """Multiple select targets. Ensure each is on a separate line.""" # Insert newline before every select target. fixes = [] for i, select_target in enumerate(select_targets_info.select_targets): base_segment = (segment if not i else select_targets_info.select_targets[i - 1]) if (base_segment.pos_marker.working_line_no == select_target.pos_marker.working_line_no): # Find and delete any whitespace before the select target. start_seg = select_targets_info.select_idx # If any select modifier (e.g. distinct ) is present, start # there rather than at the beginning. modifier = segment.get_child("select_clause_modifier") if modifier: start_seg = segment.segments.index(modifier) ws_to_delete = segment.select_children( start_seg=segment.segments[start_seg] if not i else select_targets_info.select_targets[i - 1], select_if=lambda s: s.is_type("whitespace"), loop_while=lambda s: s.is_type("whitespace", "comma") or s. is_meta, ) fixes += [LintFix.delete(ws) for ws in ws_to_delete] fixes.append( LintFix.create_before(select_target, [NewlineSegment()])) # If we are at the last select target check if the FROM clause # is on the same line, and if so move it to its own line. if select_targets_info.from_segment: if (i + 1 == len(select_targets_info.select_targets)) and ( select_target.pos_marker.working_line_no == select_targets_info.from_segment.pos_marker. working_line_no): fixes.extend([ LintFix.delete(ws) for ws in select_targets_info.pre_from_whitespace ]) fixes.append( LintFix.create_before( select_targets_info.from_segment, [NewlineSegment()], )) if fixes: return LintResult(anchor=segment, fixes=fixes)
def _eval(self, context: RuleContext) -> LintResult: """Function name not immediately followed by bracket. Look for Function Segment with anything other than the function name before brackets """ segment = context.functional.segment # We only trigger on start_bracket (open parenthesis) if segment.all(sp.is_type("function")): children = segment.children() function_name = children.first(sp.is_type("function_name"))[0] start_bracket = children.first(sp.is_type("bracketed"))[0] intermediate_segments = children.select(start_seg=function_name, stop_seg=start_bracket) if intermediate_segments: # It's only safe to fix if there is only whitespace # or newlines in the intervening section. if intermediate_segments.all( sp.is_type("whitespace", "newline")): return LintResult( anchor=intermediate_segments[0], fixes=[ LintFix.delete(seg) for seg in intermediate_segments ], ) else: # It's not all whitespace, just report the error. return LintResult(anchor=intermediate_segments[0], ) return LintResult()
def _eval(self, context: RuleContext) -> Optional[LintResult]: """Trailing commas within select clause.""" # Config type hints self.select_clause_trailing_comma: str segment = context.functional.segment children = segment.children() if segment.all(sp.is_type("select_clause")): # Iterate content to find last element last_content: BaseSegment = children.last(sp.is_code())[0] # What mode are we in? if self.select_clause_trailing_comma == "forbid": # Is it a comma? if last_content.is_type("comma"): return LintResult( anchor=last_content, fixes=[LintFix.delete(last_content)], description= "Trailing comma in select statement forbidden", ) elif self.select_clause_trailing_comma == "require": if not last_content.is_type("comma"): new_comma = SymbolSegment(",", name="comma", type="comma") return LintResult( anchor=last_content, fixes=[ LintFix.replace(last_content, [last_content, new_comma]) ], description= "Trailing comma in select statement required", ) return None
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 end with a single trailing newline. We only care about the segment and the siblings which come after it for this rule, we discard the others into the kwargs argument. """ # We only care about the final segment of the parse tree. if not self.is_final_segment(context): return None # Include current segment for complete stack and reverse. parent_stack: Segments = context.functional.parent_stack complete_stack: Segments = (context.functional.raw_stack + context.functional.segment) reversed_complete_stack = complete_stack.reversed() # Find the trailing newline segments. trailing_newlines = reversed_complete_stack.select( select_if=sp.is_type("newline"), loop_while=sp.or_(sp.is_whitespace(), sp.is_type("dedent")), ) trailing_literal_newlines = trailing_newlines.select( loop_while=lambda seg: sp.templated_slices( seg, context.templated_file).all(tsp.is_slice_type("literal"))) if not trailing_literal_newlines: # We make an edit to create this segment after the child of the FileSegment. if len(parent_stack) == 1: fix_anchor_segment = context.segment else: fix_anchor_segment = parent_stack[1] return LintResult( anchor=context.segment, fixes=[ LintFix.create_after( fix_anchor_segment, [NewlineSegment()], ) ], ) elif len(trailing_literal_newlines) > 1: # Delete extra newlines. return LintResult( anchor=context.segment, fixes=[ LintFix.delete(d) for d in trailing_literal_newlines[1:] ], ) else: # Single newline, no need for fix. return None
def _eval(self, context: RuleContext) -> Optional[LintResult]: """Commas should not have whitespace directly before them. We need at least one segment behind us for this to work. """ if len(context.raw_stack) >= 1: cm1 = context.raw_stack[-1] if (context.segment.is_type("comma") and cm1.is_type("whitespace") and cm1.pos_marker.line_pos > 1): anchor = cm1 return LintResult(anchor=anchor, fixes=[LintFix.delete(cm1)]) # Otherwise fine return None
def _coerce_indent_to( self, desired_indent: str, current_indent_buffer: Tuple[RawSegment, ...], current_anchor: BaseSegment, ) -> List[LintFix]: """Generate fixes to make an indent a certain size.""" # If there shouldn't be an indent at all, just delete. if len(desired_indent) == 0: fixes = [LintFix.delete(elem) for elem in current_indent_buffer] # If we don't have any indent and we should, then add a single elif len("".join(elem.raw for elem in current_indent_buffer)) == 0: fixes = [ LintFix.create_before( current_anchor, [ WhitespaceSegment( raw=desired_indent, ), ], ), ] # Otherwise edit the first element to be the right size else: # Edit the first element of this line's indent and remove any other # indents. fixes = [ LintFix.replace( current_indent_buffer[0], [ WhitespaceSegment( raw=desired_indent, ), ], ), ] + [LintFix.delete(elem) for elem in current_indent_buffer[1:]] return fixes
def _eval(self, context: RuleContext) -> Optional[List[LintResult]]: """Single whitespace expected in mother middle segment.""" error_buffer: List[LintResult] = [] if context.segment.is_type(self.expected_mother_segment_type): last_code = None mid_segs: List[BaseSegment] = [] for seg in context.segment.iter_segments( expanding=self.expand_children): if seg.is_code: if (last_code and self.matches_target_tuples( last_code, [self.pre_segment_identifier]) and self.matches_target_tuples( seg, [self.post_segment_identifier])): # Do we actually have the right amount of whitespace? raw_inner = "".join(s.raw for s in mid_segs) if raw_inner != " " and not (self.allow_newline and any(s.name == "newline" for s in mid_segs)): if not raw_inner.strip(): # There's some whitespace and/or newlines, or nothing fixes = [] if raw_inner: # There's whitespace and/or newlines. Drop those. fixes += [ LintFix.delete(mid_seg) for mid_seg in mid_segs ] # Enforce a single space fixes += [ LintFix.create_before( seg, [WhitespaceSegment()], ) ] else: # Don't otherwise suggest a fix for now. # Only whitespace & newlines are covered. # At least a comment section between `AS` and `(` can # result in an unfixable error. # TODO: Enable more complex fixing here. fixes = None # pragma: no cover error_buffer.append( LintResult(anchor=last_code, fixes=fixes)) mid_segs = [] if not seg.is_meta: last_code = seg else: mid_segs.append(seg) return error_buffer or None
def _handle_semicolon_same_line( context: RuleContext, info: SegmentMoveContext) -> Optional[LintResult]: if not info.before_segment: return None # If preceding segments are found then delete the old # semi-colon and its preceding whitespace and then insert # the semi-colon in the correct location. fixes = [ LintFix.replace( info.anchor_segment, [ info.anchor_segment, SymbolSegment(raw=";", type="symbol", name="semicolon"), ], ), LintFix.delete(context.segment, ), ] fixes.extend(LintFix.delete(d) for d in info.whitespace_deletions) return LintResult( anchor=info.anchor_segment, fixes=fixes, )
def _fixes_for_move_after_select_clause( stop_seg: BaseSegment, add_newline: bool = True, ) -> List[LintFix]: """Cleans up by moving leftover select_clause segments. Context: Some of the other fixes we make in _eval_single_select_target_element() leave the select_clause in an illegal state -- a select_clause's *rightmost children cannot be whitespace or comments*. This function addresses that by moving these segments up the parse tree to an ancestor segment chosen by _choose_anchor_segment(). After these fixes are applied, these segments may, for example, be *siblings* of select_clause. """ start_seg = ( modifier[0] if modifier else select_children[ select_targets_info.first_new_line_idx]) move_after_select_clause = select_children.select( start_seg=start_seg, stop_seg=stop_seg, ) fixes = [ LintFix.delete(seg) for seg in move_after_select_clause ] fixes.append( LintFix.create_after( self._choose_anchor_segment( context, "create_after", select_clause[0], filter_meta=True, ), ([NewlineSegment()] if add_newline else []) + list(move_after_select_clause), )) return fixes
def _eval(self, context: RuleContext) -> LintResult: """Unnecessary trailing whitespace. Look for newline segments, and then evaluate what it was preceded by. """ # We only trigger on newlines if (context.segment.is_type("newline") and len(context.raw_stack) > 0 and context.raw_stack[-1].is_type("whitespace")): # If we find a newline, which is preceded by whitespace, then bad deletions = context.functional.raw_stack.reversed().select( loop_while=sp.is_type("whitespace")) last_deletion_slice = deletions[-1].pos_marker.source_slice # Check the raw source (before template expansion) immediately # following the whitespace we want to delete. Often, what looks # like trailing whitespace in rendered SQL is actually a line like: # " {% for elem in elements %}\n", in which case the code is # fine -- it's not trailing whitespace from a source code # perspective. if context.templated_file: next_raw_slice = ( context.templated_file.raw_slices_spanning_source_slice( slice(last_deletion_slice.stop, last_deletion_slice.stop))) # If the next slice is literal, that means it's regular code, so # it's safe to delete the trailing whitespace. If it's anything # else, it's template code, so don't delete the whitespace because # it's not REALLY trailing whitespace in terms of the raw source # code. if next_raw_slice and next_raw_slice[0].slice_type != "literal": return LintResult() return LintResult( anchor=deletions[-1], fixes=[LintFix.delete(d) for d in deletions], ) return LintResult()
def _generate_fixes( operator_new_lines: str, change_list: Segments, operator: BaseSegment, insert_anchor: BaseSegment, ) -> LintResult: # Duplicate the change list and append the operator inserts: List[BaseSegment] = [ *change_list, operator, ] if operator_new_lines == "before": # We do yet another reverse here, # This could be avoided but makes all "changes" relate to "before" config state inserts = [*reversed(inserts)] # ensure to insert in the right place edit_type = "create_before" if operator_new_lines == "before" else "create_after" fixes = [ # Insert elements reversed LintFix( edit_type=edit_type, edit=map(lambda el: copy.deepcopy(el), reversed(inserts)), anchor=insert_anchor, ), # remove the Op LintFix.delete(operator), # Delete the original elements (related to insert) *change_list.apply(LintFix.delete), ] desc = before_description if operator_new_lines == "before" else after_description return LintResult( anchor=operator, description=desc, fixes=fixes, )
def _eval(self, context: RuleContext) -> Optional[LintResult]: """Files must not begin with newlines or whitespace.""" # If parent_stack is empty we are currently at FileSegment. if len(context.parent_stack) == 0: return None # If raw_stack is empty there can be nothing to remove. if len(context.raw_stack) == 0: return None segment = context.functional.segment raw_stack = context.functional.raw_stack whitespace_types = {"newline", "whitespace", "indent", "dedent"} # Non-whitespace segment. if ( # Non-whitespace segment. not segment.all(sp.is_type(*whitespace_types)) # We want first Non-whitespace segment so # all preceding segments must be whitespace # and at least one is not meta. and raw_stack.all(sp.is_type(*whitespace_types)) and 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.parent_stack[0], fixes=[LintFix.delete(d) for d in raw_stack], ) return None
def _process_current_line( self, res: dict, memory: dict, context: RuleContext ) -> LintResult: """Checks indentation of one line of code, returning a LintResult. The _eval() function calls it for the current line of code: - When passed a newline segment (thus ending a line) - When passed the *final* segment in the entire parse tree (which may not be a newline) """ this_line_no = max(res.keys()) this_line = res.pop(this_line_no) self.logger.debug( "Evaluating line #%s. %s", this_line_no, # Don't log the line or indent buffer, it's too noisy. self._strip_buffers(this_line), ) trigger_segment = memory["trigger"] # Is this line just comments? (Disregard trailing newline if present.) check_comment_line = this_line["line_buffer"] if check_comment_line and all( seg.is_type( "whitespace", "comment", "indent" # dedent is a subtype of indent ) for seg in check_comment_line ): # Comment line, deal with it later. memory["comment_lines"].append(this_line_no) self.logger.debug(" Comment Line. #%s", this_line_no) return LintResult(memory=memory) # Is it a hanging indent? # Find last meaningful line indent. last_code_line = None for k in sorted(res.keys(), reverse=True): if any(seg.is_code for seg in res[k]["line_buffer"]): last_code_line = k break if len(res) > 0 and last_code_line: last_line_hanger_indent = res[last_code_line]["hanging_indent"] # Let's just deal with hanging indents here. if ( # NB: Hangers are only allowed if there was content after the last # indent on the previous line. Otherwise it's just an indent. this_line["indent_size"] == last_line_hanger_indent # Or they're if the indent balance is the same and the indent is the # same AND the previous line was a hanger or ( this_line["indent_size"] == res[last_code_line]["indent_size"] and this_line["indent_balance"] == res[last_code_line]["indent_balance"] and last_code_line in memory["hanging_lines"] ) ) and ( # There MUST also be a non-zero indent. Otherwise we're just on the # baseline. this_line["indent_size"] > 0 ): # This is a HANGER memory["hanging_lines"].append(this_line_no) self.logger.debug(" Hanger Line. #%s", this_line_no) self.logger.debug( " Last Line: %s", self._strip_buffers(res[last_code_line]) ) return LintResult(memory=memory) # Is this an indented first line? elif len(res) == 0: if this_line["indent_size"] > 0: self.logger.debug(" Indented First Line. #%s", this_line_no) return LintResult( anchor=trigger_segment, memory=memory, description="First line has unexpected indent", fixes=[LintFix.delete(elem) for elem in this_line["indent_buffer"]], ) # Special handling for template end blocks on a line by themselves. if self._is_template_block_end_line( this_line["line_buffer"], context.templated_file ): block_lines = { k: ( "end" if self._is_template_block_end_line( res[k]["line_buffer"], context.templated_file ) else "start", res[k]["indent_balance"], "".join( seg.raw or getattr(seg, "source_str", "") for seg in res[k]["line_buffer"] ), ) for k in res if self._is_template_block_end_line( res[k]["line_buffer"], context.templated_file ) or self._is_template_block_start_line( res[k]["line_buffer"], context.templated_file ) } # For a template block end on a line by itself, search for a # matching block start on a line by itself. If there is one, match # its indentation. Question: Could we avoid treating this as a # special case? It has some similarities to the non-templated test # case test/fixtures/linter/indentation_error_contained.sql, in tha # both have lines where indent_balance drops 2 levels from one line # to the next, making it a bit unclear how to indent that line. template_block_level = -1 for k in sorted(block_lines.keys(), reverse=True): if block_lines[k][0] == "end": template_block_level -= 1 else: template_block_level += 1 if template_block_level != 0: continue # Found prior template block line with the same indent balance. # Is this a problem line? if k in memory["problem_lines"] + memory["hanging_lines"]: # Skip it if it is return LintResult(memory=memory) self.logger.debug(" [template block end] Comparing to #%s", k) if this_line["indent_size"] == res[k]["indent_size"]: # All good. return LintResult(memory=memory) # Indents don't match even though balance is the same... memory["problem_lines"].append(this_line_no) # The previous indent. desired_indent = "".join(elem.raw for elem in res[k]["indent_buffer"]) # Make fixes fixes = self._coerce_indent_to( desired_indent=desired_indent, current_indent_buffer=this_line["indent_buffer"], current_anchor=this_line["line_buffer"][0], ) self.logger.debug( " !! Indentation does not match #%s. Fixes: %s", k, fixes ) return LintResult( anchor=trigger_segment, memory=memory, description="Indentation not consistent with line #{}".format(k), # See above for logic fixes=fixes, ) # Assuming it's not a hanger, let's compare it to the other previous # lines. We do it in reverse so that closer lines are more relevant. for k in sorted(res.keys(), reverse=True): # Is this a problem line? if k in memory["problem_lines"] + memory["hanging_lines"]: # Skip it if it is continue # Is this an empty line? if not any( elem.is_code or elem.is_type("placeholder") for elem in res[k]["line_buffer"] ): # Skip if it is continue # Work out the difference in indent indent_diff = this_line["indent_balance"] - res[k]["indent_balance"] # If we're comparing to a previous, more deeply indented line, then skip and # keep looking. if indent_diff < 0: continue # Is the indent balance the same? if indent_diff == 0: self.logger.debug(" [same indent balance] Comparing to #%s", k) if this_line["indent_size"] != res[k]["indent_size"]: # Indents don't match even though balance is the same... memory["problem_lines"].append(this_line_no) # Work out desired indent if res[k]["indent_size"] == 0: desired_indent = "" elif this_line["indent_size"] == 0: desired_indent = self._make_indent( indent_unit=self.indent_unit, tab_space_size=self.tab_space_size, ) else: # The previous indent. desired_indent = "".join( elem.raw for elem in res[k]["indent_buffer"] ) # Make fixes fixes = self._coerce_indent_to( desired_indent=desired_indent, current_indent_buffer=this_line["indent_buffer"], current_anchor=trigger_segment, ) self.logger.debug( " !! Indentation does not match #%s. Fixes: %s", k, fixes ) return LintResult( anchor=trigger_segment, memory=memory, description="Indentation not consistent with line #{}".format( k ), # See above for logic fixes=fixes, ) # Are we at a deeper indent? elif indent_diff > 0: self.logger.debug(" [deeper indent balance] Comparing to #%s", k) # NB: We shouldn't need to deal with correct hanging indents # here, they should already have been dealt with before. We # may still need to deal with *creating* hanging indents if # appropriate. self.logger.debug( " Comparison Line: %s", self._strip_buffers(res[k]) ) # Check to see if we've got a whole number of multiples. If # we do then record the number for later, otherwise raise # an error. We do the comparison here so we have a reference # point to do the repairs. We need a sensible previous line # to base the repairs off. If there's no indent at all, then # we should also take this route because there SHOULD be one. if this_line["indent_size"] % self.tab_space_size != 0: memory["problem_lines"].append(this_line_no) # The default indent is the one just reconstructs it from # the indent size. default_indent = "".join( elem.raw for elem in res[k]["indent_buffer"] ) + self._make_indent( indent_unit=self.indent_unit, tab_space_size=self.tab_space_size, num=indent_diff, ) # If we have a clean indent, we can just add steps in line # with the difference in the indent buffers. simples. if this_line["clean_indent"]: self.logger.debug(" Use clean indent.") desired_indent = default_indent # If we have the option of a hanging indent then use it. elif res[k]["hanging_indent"]: self.logger.debug(" Use hanging indent.") desired_indent = " " * res[k]["hanging_indent"] else: # pragma: no cover self.logger.debug(" Use default indent.") desired_indent = default_indent # Make fixes fixes = self._coerce_indent_to( desired_indent=desired_indent, current_indent_buffer=this_line["indent_buffer"], current_anchor=trigger_segment, ) return LintResult( anchor=trigger_segment, memory=memory, description=( "Indentation not hanging or a multiple of {} spaces" ).format(self.tab_space_size), fixes=fixes, ) else: # We'll need this value later. this_indent_num = this_line["indent_size"] // self.tab_space_size # We know that the indent balance is higher, what actually is # the difference in indent counts? It should be a whole number # if we're still here. comp_indent_num = res[k]["indent_size"] // self.tab_space_size # The indent number should be at least 1, and can be UP TO # and including the difference in the indent balance. if comp_indent_num == this_indent_num: # We have two lines indented the same, but with a different starting # indent balance. This is either a problem OR a sign that one of the # opening indents wasn't used. We account for the latter and then # have a violation if that wasn't the case. # Does the comparison line have enough unused indent to get us back # to where we need to be? NB: This should only be applied if this is # a CLOSING bracket. # First work out if we have some closing brackets, and if so, how # many. b_idx = 0 b_num = 0 while True: if len(this_line["line_buffer"][b_idx:]) == 0: break elem = this_line["line_buffer"][b_idx] if not elem.is_code: b_idx += 1 continue else: if elem.is_type("end_bracket", "end_square_bracket"): b_idx += 1 b_num += 1 continue break # pragma: no cover if b_num >= indent_diff: # It does. This line is fine. pass else: # It doesn't. That means we *should* have an indent when # compared to this line and we DON'T. memory["problem_lines"].append(this_line_no) return LintResult( anchor=trigger_segment, memory=memory, description="Indent expected and not found compared to line" " #{}".format(k), # Add in an extra bit of whitespace for the indent fixes=[ LintFix.create_before( trigger_segment, [ WhitespaceSegment( raw=self._make_indent( indent_unit=self.indent_unit, tab_space_size=self.tab_space_size, ), ), ], ), ], ) elif this_indent_num < comp_indent_num: memory["problem_lines"].append(this_line_no) return LintResult( anchor=trigger_segment, memory=memory, description="Line under-indented compared to line #{}".format( k ), fixes=[ LintFix.create_before( trigger_segment, [ WhitespaceSegment( # Make the minimum indent for it to be ok. raw=self._make_indent( num=comp_indent_num - this_indent_num, indent_unit=self.indent_unit, tab_space_size=self.tab_space_size, ), ), ], ), ], ) elif this_indent_num > comp_indent_num + indent_diff: # Calculate the lowest ok indent: desired_indent = self._make_indent( num=comp_indent_num - this_indent_num, indent_unit=self.indent_unit, tab_space_size=self.tab_space_size, ) # Make fixes fixes = self._coerce_indent_to( desired_indent=desired_indent, current_indent_buffer=this_line["indent_buffer"], current_anchor=trigger_segment, ) memory["problem_lines"].append(this_line_no) return LintResult( anchor=trigger_segment, memory=memory, description="Line over-indented compared to line #{}".format(k), fixes=fixes, ) # This was a valid comparison, so if it doesn't flag then # we can assume that we're ok. self.logger.debug(" Indent deemed ok comparing to #%s", k) # Given that this line is ok, consider if the preceding lines are # comments. If they are, lint the indentation of the comment(s). fixes = [] for n in range(this_line_no - 1, -1, -1): if n in memory["comment_lines"]: # The previous line WAS a comment. prev_line = res[n] if this_line["indent_size"] != prev_line["indent_size"]: # It's not aligned. # Find the anchor first. anchor: BaseSegment = None # type: ignore for seg in prev_line["line_buffer"]: if seg.is_type("comment"): anchor = seg break # Make fixes. fixes += self._coerce_indent_to( desired_indent="".join( elem.raw for elem in this_line["indent_buffer"] ), current_indent_buffer=prev_line["indent_buffer"], current_anchor=anchor, ) memory["problem_lines"].append(n) else: break if fixes: return LintResult( anchor=anchor, memory=memory, description="Comment not aligned with following line.", fixes=fixes, ) # Otherwise all good. return LintResult(memory=memory) # NB: At shallower indents, we don't check, we just check the # previous lines with the same balance. Deeper indents can check # themselves. # If we get to here, then we're all good for now. return LintResult(memory=memory)
def _lint_aliases_in_join(self, base_table, from_expression_elements, column_reference_segments, segment): """Lint and fix all aliases in joins - except for self-joins.""" # A buffer to keep any violations. violation_buff = [] to_check = list( self._filter_table_expressions(base_table, from_expression_elements)) # How many times does each table appear in the FROM clause? table_counts = Counter(ai.table_ref.raw for ai in to_check) # What is the set of aliases used for each table? (We are mainly # interested in the NUMBER of different aliases used.) table_aliases = defaultdict(set) for ai in to_check: if ai and ai.table_ref and ai.alias_identifier_ref: table_aliases[ai.table_ref.raw].add( ai.alias_identifier_ref.raw) # For each aliased table, check whether to keep or remove it. for alias_info in to_check: # If the same table appears more than once in the FROM clause with # different alias names, do not consider removing its aliases. # The aliases may have been introduced simply to make each # occurrence of the table independent within the query. if (table_counts[alias_info.table_ref.raw] > 1 and len(table_aliases[alias_info.table_ref.raw]) > 1): continue select_clause = segment.get_child("select_clause") ids_refs = [] # Find all references to alias in select clause if alias_info.alias_identifier_ref: alias_name = alias_info.alias_identifier_ref.raw for alias_with_column in select_clause.recursive_crawl( "object_reference"): used_alias_ref = alias_with_column.get_child("identifier") if used_alias_ref and used_alias_ref.raw == alias_name: ids_refs.append(used_alias_ref) # Find all references to alias in column references for exp_ref in column_reference_segments: used_alias_ref = exp_ref.get_child("identifier") # exp_ref.get_child('dot') ensures that the column reference includes a # table reference if (used_alias_ref and used_alias_ref.raw == alias_name and exp_ref.get_child("dot")): ids_refs.append(used_alias_ref) # Fixes for deleting ` as sth` and for editing references to aliased tables # Note unparsable errors have cause the delete to fail (see #2484) # so check there is a d before doing deletes. fixes = [ *[ LintFix.delete(d) for d in [alias_info.alias_exp_ref, alias_info.whitespace_ref] if d ], *[ LintFix.replace( alias, [alias.edit(alias_info.table_ref.raw)], source=[alias_info.table_ref], ) for alias in [alias_info.alias_identifier_ref, *ids_refs] if alias ], ] violation_buff.append( LintResult( anchor=alias_info.alias_identifier_ref, description= "Avoid aliases in from clauses and join conditions.", fixes=fixes, )) return violation_buff or None
def _eval(self, context: RuleContext) -> LintResult: """WITH clause closing bracket should be aligned with WITH keyword. Look for a with clause and evaluate the position of closing brackets. """ # We only trigger on start_bracket (open parenthesis) if context.segment.is_type("with_compound_statement"): raw_stack_buff = list(context.raw_stack) # Look for the with keyword for seg in context.segment.segments: if seg.name.lower() == "with": seg_line_no = seg.pos_marker.line_no break else: # pragma: no cover # This *could* happen if the with statement is unparsable, # in which case then the user will have to fix that first. if any( s.is_type("unparsable") for s in context.segment.segments): return LintResult() # If it's parsable but we still didn't find a with, then # we should raise that. raise RuntimeError("Didn't find WITH keyword!") def indent_size_up_to(segs): seg_buff = [] # Get any segments running up to the WITH for elem in reversed(segs): if elem.is_type("newline"): break elif elem.is_meta: continue else: seg_buff.append(elem) # reverse the indent if we have one if seg_buff: seg_buff = list(reversed(seg_buff)) indent_str = "".join(seg.raw for seg in seg_buff).replace( "\t", " " * self.tab_space_size) indent_size = len(indent_str) return indent_size, indent_str balance = 0 with_indent, with_indent_str = indent_size_up_to(raw_stack_buff) for seg in context.segment.iter_segments( expanding=["common_table_expression", "bracketed"], pass_through=True): if seg.name == "start_bracket": balance += 1 elif seg.name == "end_bracket": balance -= 1 if balance == 0: closing_bracket_indent, _ = indent_size_up_to( raw_stack_buff) indent_diff = closing_bracket_indent - with_indent # Is indent of closing bracket not the same as # indent of WITH keyword. if seg.pos_marker.line_no == seg_line_no: # Skip if it's the one-line version. That's ok pass elif indent_diff < 0: return LintResult( anchor=seg, fixes=[ LintFix.create_before( seg, [ WhitespaceSegment(" " * (-indent_diff)) ], ) ], ) elif indent_diff > 0: # Is it all whitespace before the bracket on this line? assert seg.pos_marker prev_segs_on_line = [ elem for elem in context.segment.raw_segments if cast(PositionMarker, elem.pos_marker ).line_no == seg.pos_marker.line_no and cast(PositionMarker, elem.pos_marker ).line_pos < seg.pos_marker.line_pos ] if all( elem.is_type("whitespace") for elem in prev_segs_on_line): # We can move it back, it's all whitespace fixes = [ LintFix.create_before( seg, [WhitespaceSegment(with_indent_str)], ) ] + [ LintFix.delete(elem) for elem in prev_segs_on_line ] else: # We have to move it to a newline fixes = [ LintFix.create_before( seg, [ NewlineSegment(), WhitespaceSegment(with_indent_str), ], ) ] return LintResult(anchor=seg, fixes=fixes) else: raw_stack_buff.append(seg) return LintResult()
def _eval(self, context: RuleContext) -> Optional[LintResult]: """Line is too long. This only triggers on newline segments, evaluating the whole line. The detection is simple, the fixing is much trickier. """ # Config type hints self.max_line_length: int self.ignore_comment_lines: bool self.ignore_comment_clauses: bool if not context.memory: memory: dict = {"comment_clauses": set()} else: memory = context.memory if context.segment.name == "newline": # iterate to buffer the whole line up to this point this_line = self._gen_line_so_far(context.raw_stack) else: if self.ignore_comment_clauses and context.segment.is_type( "comment_clause", "comment_equals_clause"): comment_segment = context.functional.segment.children().first( sp.is_name("quoted_literal")) if comment_segment: memory["comment_clauses"].add(comment_segment.get()) # Otherwise we're all good return LintResult(memory=memory) # Now we can work out the line length and deal with the content line_len = self._compute_source_length(this_line, memory) if line_len > self.max_line_length: # Problem, we'll be reporting a violation. The # question is, can we fix it? # We'll need the indent, so let's get it for fixing. line_indent = [] for s in this_line: if s.name == "whitespace": line_indent.append(s) else: break # Don't even attempt to handle template placeholders as gets # complicated if logic changes (e.g. moving for loops). Most of # these long lines will likely be single line Jinja comments. # They will remain as unfixable. if this_line[-1].type == "placeholder": self.logger.info("Unfixable template segment: %s", this_line[-1]) return LintResult(anchor=context.segment, memory=memory) # Does the line end in an inline comment that we can move back? if this_line[-1].name == "inline_comment": # Is this line JUST COMMENT (with optional preceding whitespace) if # so, user will have to fix themselves. if len(this_line) == 1 or all( elem.name == "whitespace" or elem.is_meta for elem in this_line[:-1]): self.logger.info( "Unfixable inline comment, alone on line: %s", this_line[-1]) if self.ignore_comment_lines: return LintResult(memory=memory) else: return LintResult(anchor=context.segment, memory=memory) self.logger.info( "Attempting move of inline comment at end of line: %s", this_line[-1], ) # Set up to delete the original comment and the preceding whitespace delete_buffer = [LintFix.delete(this_line[-1])] idx = -2 while True: if (len(this_line) >= abs(idx) and this_line[idx].name == "whitespace"): delete_buffer.append(LintFix.delete(this_line[idx])) idx -= 1 else: break # pragma: no cover create_elements = line_indent + [ this_line[-1], cast(RawSegment, context.segment), ] if (self._compute_source_length(create_elements, memory) > self.max_line_length): # The inline comment is NOT on a line by itself, but even if # we move it onto a line by itself, it's still too long. In # this case, the rule should do nothing, otherwise it # triggers an endless cycle of "fixes" that simply keeps # adding blank lines. self.logger.info( "Unfixable inline comment, too long even on a line by itself: " "%s", this_line[-1], ) if self.ignore_comment_lines: return LintResult(memory=memory) else: return LintResult(anchor=context.segment, memory=memory) # Create a newline before this one with the existing comment, an # identical indent AND a terminating newline, copied from the current # target segment. create_buffer = [ LintFix.create_before(this_line[0], create_elements) ] return LintResult( anchor=context.segment, fixes=delete_buffer + create_buffer, memory=memory, ) fixes = self._eval_line_for_breaks(this_line) if fixes: return LintResult(anchor=context.segment, fixes=fixes, memory=memory) return LintResult(anchor=context.segment, memory=memory) return LintResult(memory=memory)
def _eval(self, context: RuleContext) -> Optional[LintResult]: """Select clause modifiers must appear on same line as SELECT.""" # We only care about select_clause. if not context.segment.is_type("select_clause"): return None # Get children of select_clause and the corresponding select keyword. child_segments = context.functional.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 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 between select keyword and select clause modifier # and delete the original select clause modifier. fixes.extend((LintFix.delete(s) for s in leading_newline_segments)) 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, )
def _eval(self, context: RuleContext) -> Optional[List[LintResult]]: """Unnecessary whitespace.""" # For the given segment, lint whitespace directly within it. prev_newline = True prev_whitespace = None violations = [] for seg in context.segment.segments: if seg.is_meta: continue elif seg.is_type("newline"): prev_newline = True prev_whitespace = None elif seg.is_type("whitespace"): # This is to avoid indents if not prev_newline: prev_whitespace = seg # We won't set prev_newline to False, just for whitespace # in case there's multiple indents, inserted by other rule # fixes (see #1713) elif seg.is_type("comment"): prev_newline = False prev_whitespace = None else: if prev_whitespace: if prev_whitespace.raw != " ": violations.append( LintResult( anchor=prev_whitespace, fixes=[ LintFix.replace( prev_whitespace, [WhitespaceSegment()], ) ], )) prev_newline = False prev_whitespace = None if seg.is_type("object_reference"): for child_seg in seg.get_raw_segments(): if child_seg.is_whitespace: violations.append( LintResult( anchor=child_seg, fixes=[LintFix.delete(child_seg)], )) if seg.is_type("comparison_operator"): delete_fixes = [ LintFix.delete(s) for s in seg.get_raw_segments() if s.is_whitespace ] if delete_fixes: violations.append( LintResult( anchor=delete_fixes[0].anchor, fixes=delete_fixes, )) if context.segment.is_type("casting_operator"): leading_whitespace_segments = ( context.functional.raw_stack.reversed().select( select_if=sp.is_whitespace(), loop_while=sp.or_(sp.is_whitespace(), sp.is_meta()), )) trailing_whitespace_segments = ( context.functional.siblings_post.raw_segments.select( select_if=sp.is_whitespace(), loop_while=sp.or_(sp.is_whitespace(), sp.is_meta()), )) fixes: List[LintFix] = [] fixes.extend( LintFix.delete(s) for s in leading_whitespace_segments) fixes.extend( LintFix.delete(s) for s in trailing_whitespace_segments) if fixes: violations.append( LintResult( anchor=context.segment, fixes=fixes, )) return violations or None
def _validate_one_reference( single_table_references: str, allow_select_alias: bool, ref: BaseSegment, this_ref_type: str, standalone_aliases: List[str], table_ref_str: str, col_alias_names: List[str], seen_ref_types: Set[str], ) -> Optional[LintResult]: # We skip any unqualified wildcard references (i.e. *). They shouldn't # count. if not ref.is_qualified() and ref.is_type( "wildcard_identifier"): # type: ignore return None # Oddball case: Column aliases provided via function calls in by # FROM or JOIN. References to these don't need to be qualified. # Note there could be a table with a column by the same name as # this alias, so avoid bogus warnings by just skipping them # entirely rather than trying to enforce anything. if ref.raw in standalone_aliases: return None # Certain dialects allow use of SELECT alias in WHERE clauses if allow_select_alias and ref.raw in col_alias_names: return None if single_table_references == "consistent": if seen_ref_types and this_ref_type not in seen_ref_types: return LintResult( anchor=ref, description=f"{this_ref_type.capitalize()} reference " f"{ref.raw!r} found in single table select which is " "inconsistent with previous references.", ) return None if single_table_references != this_ref_type: if single_table_references == "unqualified": # If this is qualified we must have a "table", "."" at least fixes = [LintFix.delete(el) for el in ref.segments[:2]] return LintResult( anchor=ref, fixes=fixes, description="{} reference {!r} found in single table " "select.".format(this_ref_type.capitalize(), ref.raw), ) fixes = [ LintFix.create_before( ref.segments[0] if len(ref.segments) else ref, edit_segments=[ CodeSegment( raw=table_ref_str, name="naked_identifier", type="identifier", ), SymbolSegment(raw=".", type="symbol", name="dot"), ], ) ] return LintResult( anchor=ref, fixes=fixes, description="{} reference {!r} found in single table " "select.".format(this_ref_type.capitalize(), ref.raw), ) return None
def _eval(self, context: RuleContext) -> Optional[LintResult]: """Implicit aliasing of table/column not allowed. Use explicit `AS` clause. We look for the alias segment, and then evaluate its parent and whether it contains an AS keyword. This is the _eval function for both L011 and L012. The use of `raw_stack` is just for working out how much whitespace to add. """ # Config type hints self.aliasing: str fixes = [] if context.segment.is_type("alias_expression"): if context.parent_stack[-1].is_type(*self._target_elems): if any(e.name.lower() == "as" for e in context.segment.segments): if self.aliasing == "implicit": if context.segment.segments[0].name.lower() == "as": # Remove the AS as we're using implict aliasing fixes.append( LintFix.delete(context.segment.segments[0])) anchor = context.raw_stack[-1] # Remove whitespace before (if exists) or after (if not) if (len(context.raw_stack) > 0 and context.raw_stack[-1].type == "whitespace"): fixes.append( LintFix.delete(context.raw_stack[-1])) elif (len(context.segment.segments) > 0 and context.segment.segments[1].type == "whitespace"): fixes.append( LintFix.delete( context.segment.segments[1])) return LintResult(anchor=anchor, fixes=fixes) elif self.aliasing != "implicit": insert_buff: List[Union[WhitespaceSegment, KeywordSegment]] = [] # Add initial whitespace if we need to... if context.raw_stack[-1].name not in [ "whitespace", "newline" ]: insert_buff.append(WhitespaceSegment()) # Add an AS (Uppercase for now, but could be corrected later) insert_buff.append(KeywordSegment("AS")) # Add a trailing whitespace if we need to if context.segment.segments[0].name not in [ "whitespace", "newline", ]: insert_buff.append(WhitespaceSegment()) return LintResult( anchor=context.segment, fixes=[ LintFix.create_before( context.segment.segments[0], insert_buff, ) ], ) return None
def _eval(self, context: RuleContext) -> Optional[LintResult]: """Look for USING in a join clause.""" segment = context.functional.segment parent_stack = context.functional.parent_stack # We are not concerned with non join clauses if not segment.all(sp.is_type("join_clause")): return None using_anchor = segment.children(sp.is_keyword("using")).first() # If there is no evidence of a USING then we exit if len(using_anchor) == 0: return None anchor = using_anchor.get() description = "Found USING statement. Expected only ON statements." # All returns from here out will be some form of linting error. # we prepare the variable here unfixable_result = LintResult( anchor=anchor, description=description, ) tables_in_join = parent_stack.last().children( sp.is_type("join_clause", "from_expression_element")) # If we have more than 2 tables we won't try to fix join. # TODO: if this is table 2 of 3 it is still fixable if len(tables_in_join) > 2: return unfixable_result parent_select = parent_stack.last(sp.is_type("select_statement")).get() if not parent_select: # pragma: no cover return unfixable_result select_info = get_select_statement_info(parent_select, context.dialect) if not select_info: # pragma: no cover return unfixable_result to_delete, insert_after_anchor = _extract_deletion_sequence_and_anchor( tables_in_join.last()) table_a, table_b = select_info.table_aliases edit_segments = [ KeywordSegment(raw="ON"), WhitespaceSegment(raw=" "), ] + _generate_join_conditions( table_a.ref_str, table_b.ref_str, select_info.using_cols, ) fixes = [ LintFix.create_before( anchor_segment=insert_after_anchor, edit_segments=edit_segments, ), *[LintFix.delete(seg) for seg in to_delete], ] return LintResult( anchor=anchor, description=description, fixes=fixes, )
def _eval(self, context: RuleContext) -> Optional[LintResult]: """Enforce comma placement. For leading commas we're looking for trailing commas, so we look for newline segments. For trailing commas we're looking for leading commas, so we look for the comma itself. We also want to handle proper whitespace removal/addition. We remove any trailing whitespace after the leading comma, when converting a leading comma to a trailing comma. We add whitespace after the leading comma when converting a trailing comma to a leading comma. """ # Config type hints self.comma_style: str if not context.memory: memory: Dict[str, Any] = { # Trailing comma keys # # Do we have a fix in place for removing a leading # comma violation, and inserting a new trailing comma? "insert_trailing_comma": False, # A list of whitespace segments that come after a # leading comma violation, to be removed during fixing. "whitespace_deletions": None, # The leading comma violation segment to be removed during fixing "last_leading_comma_seg": None, # The newline segment where we're going to insert our new trailing # comma during fixing "anchor_for_new_trailing_comma_seg": None, # # Leading comma keys # # Do we have a fix in place for removing a trailing # comma violation, and inserting a new leading comma? "insert_leading_comma": False, # The trailing comma violation segment to be removed during fixing "last_trailing_comma_segment": None, } else: memory = context.memory if self.comma_style == "trailing": # A comma preceded by a new line == a leading comma if context.segment.is_type("comma"): last_seg = self._last_code_seg(context.raw_stack) if (last_seg and last_seg.is_type("newline") and not last_seg.is_templated): # Recorded where the fix should be applied memory["last_leading_comma_seg"] = context.segment last_comment_seg = self._last_comment_seg( context.raw_stack) inline_comment = (last_comment_seg.pos_marker.line_no == last_seg.pos_marker.line_no if last_comment_seg else False) # If we have a comment right before the newline, then anchor # the fix at the comment instead memory["anchor_for_new_trailing_comma_seg"] = ( last_seg if not inline_comment else last_comment_seg) # Trigger fix routine memory["insert_trailing_comma"] = True memory["whitespace_deletions"] = [] return LintResult(memory=memory) # Have we found a leading comma violation? if memory["insert_trailing_comma"]: # Search for trailing whitespace to delete after the leading # comma violation if context.segment.is_type("whitespace"): memory["whitespace_deletions"] += [context.segment] return LintResult(memory=memory) else: # We've run out of whitespace to delete, time to fix last_leading_comma_seg = memory["last_leading_comma_seg"] # Scan backwards to find the last code segment, skipping # over lines that are either entirely blank or just a # comment. We want to place the comma immediately after it. last_code_seg = None while last_code_seg is None or last_code_seg.is_type( "newline"): last_code_seg = self._last_code_seg( context.raw_stack[:context.raw_stack.index( last_code_seg if last_code_seg else memory["last_leading_comma_seg"])]) return LintResult( anchor=last_leading_comma_seg, description= "Found leading comma. Expected only trailing.", fixes=[ LintFix.delete(last_leading_comma_seg), *[ LintFix.delete(d) for d in memory["whitespace_deletions"] ], LintFix.create_before( anchor_segment=self._get_following_seg( context.raw_stack, last_code_seg), edit_segments=[last_leading_comma_seg], ), ], ) elif self.comma_style == "leading": # A new line preceded by a comma == a trailing comma if context.segment.is_type("newline"): last_seg = self._last_code_seg(context.raw_stack) # no code precedes the current position: no issue if last_seg is None: return None if last_seg.is_type( "comma") and not context.segment.is_templated: # Trigger fix routine memory["insert_leading_comma"] = True # Record where the fix should be applied memory["last_trailing_comma_segment"] = last_seg return LintResult(memory=memory) # Have we found a trailing comma violation? if memory["insert_leading_comma"]: # Only insert the comma here if this isn't a comment/whitespace segment if context.segment.is_code: last_comma_seg = memory["last_trailing_comma_segment"] # Create whitespace to insert after the new leading comma new_whitespace_seg = WhitespaceSegment() return LintResult( anchor=last_comma_seg, description= "Found trailing comma. Expected only leading.", fixes=[ LintFix.delete(last_comma_seg), LintFix.create_before( anchor_segment=context.segment, edit_segments=[ last_comma_seg, new_whitespace_seg ], ), ], ) # Otherwise, no issue return None
def _eval_single_select_target_element(self, select_targets_info, context: RuleContext): select_clause = context.functional.segment parent_stack = context.parent_stack wildcards = select_clause.children( sp.is_type("select_clause_element")).children( sp.is_type("wildcard_expression")) is_wildcard = bool(wildcards) if is_wildcard: wildcard_select_clause_element = wildcards[0] if (select_targets_info.select_idx < select_targets_info.first_new_line_idx < select_targets_info.first_select_target_idx) and ( not is_wildcard): # Do we have a modifier? select_children = select_clause.children() modifier: Optional[Segments] modifier = select_children.first( sp.is_type("select_clause_modifier")) # Prepare the select clause which will be inserted insert_buff = [ WhitespaceSegment(), select_children[select_targets_info.first_select_target_idx], ] # Check if the modifier is one we care about if modifier: # If it's already on the first line, ignore it. if (select_children.index(modifier.get()) < select_targets_info.first_new_line_idx): modifier = None fixes = [ # Delete the first select target from its original location. # We'll add it to the right section at the end, once we know # what to add. LintFix.delete( select_children[ select_targets_info.first_select_target_idx], ), ] # If we have a modifier to move: if modifier: # Add it to the insert insert_buff = [WhitespaceSegment(), modifier[0]] + insert_buff modifier_idx = select_children.index(modifier.get()) # Delete the whitespace after it (which is two after, thanks to indent) if (len(select_children) > modifier_idx + 1 and select_children[modifier_idx + 2].is_whitespace): fixes += [ LintFix.delete(select_children[modifier_idx + 2], ), ] # Delete the modifier itself fixes += [ LintFix.delete(modifier[0], ), ] # Set the position marker for removing the preceding # whitespace and newline, which we'll use below. start_idx = modifier_idx else: # Set the position marker for removing the preceding # whitespace and newline, which we'll use below. start_idx = select_targets_info.first_select_target_idx if parent_stack and parent_stack[-1].is_type("select_statement"): select_stmt = parent_stack[-1] select_clause_idx = select_stmt.segments.index( select_clause.get()) after_select_clause_idx = select_clause_idx + 1 if len(select_stmt.segments) > after_select_clause_idx: def _fixes_for_move_after_select_clause( stop_seg: BaseSegment, add_newline: bool = True, ) -> List[LintFix]: """Cleans up by moving leftover select_clause segments. Context: Some of the other fixes we make in _eval_single_select_target_element() leave the select_clause in an illegal state -- a select_clause's *rightmost children cannot be whitespace or comments*. This function addresses that by moving these segments up the parse tree to an ancestor segment chosen by _choose_anchor_segment(). After these fixes are applied, these segments may, for example, be *siblings* of select_clause. """ start_seg = ( modifier[0] if modifier else select_children[ select_targets_info.first_new_line_idx]) move_after_select_clause = select_children.select( start_seg=start_seg, stop_seg=stop_seg, ) fixes = [ LintFix.delete(seg) for seg in move_after_select_clause ] fixes.append( LintFix.create_after( self._choose_anchor_segment( context, "create_after", select_clause[0], filter_meta=True, ), ([NewlineSegment()] if add_newline else []) + list(move_after_select_clause), )) return fixes if select_stmt.segments[after_select_clause_idx].is_type( "newline"): # Since we're deleting the newline, we should also delete all # whitespace before it or it will add random whitespace to # following statements. So walk back through the segment # deleting whitespace until you get the previous newline, or # something else. to_delete = select_children.reversed().select( loop_while=sp.is_type("whitespace"), start_seg=select_children[start_idx], ) fixes += [LintFix.delete(seg) for seg in to_delete] # The select_clause is immediately followed by a # newline. Delete the newline in order to avoid leaving # behind an empty line after fix, *unless* we stopped # due to something other than a newline. delete_last_newline = select_children[ start_idx - len(to_delete) - 1].is_type("newline") # Delete the newline if we decided to. if delete_last_newline: fixes.append( LintFix.delete( select_stmt. segments[after_select_clause_idx], )) fixes += _fixes_for_move_after_select_clause( to_delete[-1], ) elif select_stmt.segments[after_select_clause_idx].is_type( "whitespace"): # The select_clause has stuff after (most likely a comment) # Delete the whitespace immediately after the select clause # so the other stuff aligns nicely based on where the select # clause started fixes += [ LintFix.delete( select_stmt.segments[after_select_clause_idx], ), ] fixes += _fixes_for_move_after_select_clause( select_children[ select_targets_info.first_select_target_idx], ) elif select_stmt.segments[after_select_clause_idx].is_type( "dedent"): # Again let's strip back the whitespace, but simpler # as don't need to worry about new line so just break # if see non-whitespace to_delete = select_children.reversed().select( loop_while=sp.is_type("whitespace"), start_seg=select_children[select_clause_idx - 1], ) fixes += [LintFix.delete(seg) for seg in to_delete] if to_delete: fixes += _fixes_for_move_after_select_clause( to_delete[-1], # If we deleted a newline, create a newline. any(seg for seg in to_delete if seg.is_type("newline")), ) else: fixes += _fixes_for_move_after_select_clause( select_children[ select_targets_info.first_select_target_idx], ) fixes += [ # Insert the select_clause in place of the first newline in the # Select statement LintFix.replace( select_children[select_targets_info.first_new_line_idx], insert_buff, ), ] return LintResult( anchor=select_clause.get(), fixes=fixes, ) # If we have a wildcard on the same line as the FROM keyword, but not the same # line as the SELECT keyword, we need to move the FROM keyword to its own line. # i.e. # SELECT # * FROM foo if select_targets_info.from_segment: if (is_wildcard and (select_clause[0].pos_marker.working_line_no != select_targets_info.from_segment.pos_marker.working_line_no) and (wildcard_select_clause_element.pos_marker.working_line_no == select_targets_info.from_segment.pos_marker.working_line_no)): fixes = [ LintFix.delete(ws) for ws in select_targets_info.pre_from_whitespace ] fixes.append( LintFix.create_before( select_targets_info.from_segment, [NewlineSegment()], )) return LintResult(anchor=select_clause.get(), fixes=fixes) return None
def generate_fixes_to_coerce( self, segments: List[RawSegment], indent_section: "Section", crawler: Rule_L016, indent: int, ) -> List[LintFix]: """Generate a list of fixes to create a break at this point. The `segments` argument is necessary to extract anchors from the existing segments. """ fixes = [] # Generate some sample indents: unit_indent = crawler._make_indent( indent_unit=crawler.indent_unit, tab_space_size=crawler.tab_space_size, ) indent_p1 = indent_section.raw + unit_indent if unit_indent in indent_section.raw: indent_m1 = indent_section.raw.replace(unit_indent, "", 1) else: indent_m1 = indent_section.raw if indent > 0: new_indent = indent_p1 elif indent < 0: new_indent = indent_m1 else: new_indent = indent_section.raw create_anchor = self.find_segment_at( segments, self.segments[-1].get_end_loc()) if self.role == "pausepoint": # Assume that this means there isn't a breakpoint # and that we'll break with the same indent as the # existing line. # NOTE: Deal with commas and binary operators differently here. # Maybe only deal with commas to start with? if any( seg.is_type("binary_operator") for seg in self.segments): # pragma: no cover raise NotImplementedError( "Don't know how to deal with binary operators here yet!!" ) # Remove any existing whitespace for elem in self.segments: if not elem.is_meta and elem.is_type("whitespace"): fixes.append(LintFix.delete(elem)) # Create a newline and a similar indent fixes.append( LintFix.create_before( create_anchor, [ NewlineSegment(), WhitespaceSegment(new_indent), ], )) return fixes if self.role == "breakpoint": # Can we determine the required indent just from # the info in this segment only? # Remove anything which is already here for elem in self.segments: if not elem.is_meta: fixes.append(LintFix.delete(elem)) # Create a newline, create an indent of the relevant size fixes.append( LintFix.create_before( create_anchor, [ NewlineSegment(), WhitespaceSegment(new_indent), ], )) return fixes raise ValueError(f"Unexpected break generated at {self}" ) # pragma: no cover