Lexers for various SQL dialects and related interactive sessions.
Postgres specific lexers:
`PostgresLexer`
A SQL lexer for the PostgreSQL dialect. Differences w.r.t. the SQL
lexer are:
- keywords and data types list parsed from the PG docs (run the
`_postgres_builtins` module to update them);
- Content of $-strings parsed using a specific lexer, e.g. the content
of a PL/Python function is parsed using the Python lexer;
- parse PG specific constructs: E-strings, $-strings, U&-strings,
different operators and punctuation.
`PlPgsqlLexer`
A lexer for the PL/pgSQL language. Adds a few specific construct on
top of the PG SQL lexer (such as <<label>>).
`PostgresConsoleLexer`
A lexer to highlight an interactive psql session:
- identifies the prompt and does its best to detect the end of command in multiline statement where not all the lines are prefixed by a
prompt, telling them apart from the output;
- highlights errors in the output and notification levels;
- handles psql backslash commands.
`PostgresExplainLexer`
A lexer to highlight Postgres execution plan.
The ``tests/examplefiles`` contains a few test files with data to be
parsed by these lexers.
:copyright: Copyright 2006-2024 by the Pygments team, see AUTHORS.
:license: BSD, see LICENSE for details. """
import re
from pygments.lexer import Lexer, RegexLexer, do_insertions, bygroups, words from pygments.token import Punctuation, Whitespace, Text, Comment, Operator, \
Keyword, Name, String, Number, Generic, Literal from pygments.lexers import get_lexer_by_name, ClassNotFound
def language_callback(lexer, match): """Parse the content of a $-string using a lexer
The lexer is chosen looking for a nearby LANGUAGE or assumed as
plpgsql if inside a DO statement and no LANGUAGE has been found. """
lx = None
m = language_re.match(lexer.text[match.end():match.end()+100]) if m isnotNone:
lx = lexer._get_lexer(m.group(1)) else:
m = list(language_re.finditer(
lexer.text[max(0, match.start()-100):match.start()])) if m:
lx = lexer._get_lexer(m[-1].group(1)) else:
m = list(do_re.finditer(
lexer.text[max(0, match.start()-25):match.start()])) if m:
lx = lexer._get_lexer('plpgsql')
class PostgresBase: """Base class for Postgres-related lexers.
This is implemented as a mixin to avoid the Lexer metaclass kicking in.
this way the different lexer don't have a common Lexer ancestor. If they
had, _tokens could be created on this ancestor andnot updated for the
other classes, resulting e.g. in PL/pgSQL parsed as SQL. This shortcoming
seem to suggest that regexp lexers are not really subclassable. """ def get_tokens_unprocessed(self, text, *args): # Have a copy of the entire text to be used by `language_callback`.
self.text = text yieldfrom super().get_tokens_unprocessed(text, *args)
def _get_lexer(self, lang): if lang.lower() == 'sql': return get_lexer_by_name('postgresql', **self.options)
tries = [lang] if lang.startswith('pl'):
tries.append(lang[2:]) if lang.endswith('u'):
tries.append(lang[:-1]) if lang.startswith('pl') and lang.endswith('u'):
tries.append(lang[2:-1])
for lx in tries: try: return get_lexer_by_name(lx, **self.options) except ClassNotFound: pass else: # TODO: better logging # print >>sys.stderr, "language not found:", lang returnNone
class PostgresLexer(PostgresBase, RegexLexer): """
Lexer for the PostgreSQL dialect of SQL. """
class PlPgsqlLexer(PostgresBase, RegexLexer): """
Handle the extra syntax in Pl/pgSQL language. """
name = 'PL/pgSQL'
aliases = ['plpgsql']
mimetypes = ['text/x-plpgsql']
url = 'https://www.postgresql.org/docs/current/plpgsql.html'
version_added = '1.5'
flags = re.IGNORECASE # FIXME: use inheritance
tokens = {name: state[:] for (name, state) in PostgresLexer.tokens.items()}
# extend the keywords list for i, pattern in enumerate(tokens['root']): if pattern[1] == Keyword:
tokens['root'][i] = (
words(KEYWORDS + PLPGSQL_KEYWORDS, suffix=r'\b'),
Keyword) del i break else: assert 0, "SQL keywords not found"
# Add specific PL/pgSQL rules (before the SQL ones)
tokens['root'][:0] = [
(r'\%[a-z]\w*\b', Name.Builtin), # actually, a datatype
(r':=', Operator),
(r'\<\<[a-z]\w*\>\>', Name.Label),
(r'\#[a-z]\w*\b', Keyword.Pseudo), # #variable_conflict
]
class PsqlRegexLexer(PostgresBase, RegexLexer): """
Extend the PostgresLexer adding support specific for psql commands.
This isnot a complete psql lexer yet as it lacks prompt support and output rendering. """
name = 'PostgreSQL console - regexp based lexer'
aliases = [] # not public
flags = re.IGNORECASE
tokens = {name: state[:] for (name, state) in PostgresLexer.tokens.items()}
# consume the lines of the command: start with an optional prompt # and continue until the end of command is detected
curcode = ''
insertions = [] for line in lines: # Identify a shell prompt in case of psql commandline example if line.startswith('$') andnot curcode:
lexer = get_lexer_by_name('console', **self.options) yieldfrom lexer.get_tokens_unprocessed(line) break
# Identify a psql prompt
mprompt = re_prompt.match(line) if mprompt isnotNone:
insertions.append((len(curcode),
[(0, Generic.Prompt, mprompt.group())]))
curcode += line[len(mprompt.group()):] else:
curcode += line
# Check if this is the end of the command # TODO: better handle multiline comments at the end with # a lexer with an external state? if re_psql_command.match(curcode) \ or re_end_command.search(curcode): break
# Emit the combined stream of command and prompt(s) yieldfrom do_insertions(insertions,
sql.get_tokens_unprocessed(curcode))
# Emit the output lines
out_token = Generic.Output for line in lines:
mprompt = re_prompt.match(line) if mprompt isnotNone: # push the line back to have it processed by the prompt
lines.send(line) break
mmsg = re_message.match(line) if mmsg isnotNone: if mmsg.group(1).startswith("ERROR") \ or mmsg.group(1).startswith("FATAL"):
out_token = Generic.Error yield (mmsg.start(1), Generic.Strong, mmsg.group(1)) yield (mmsg.start(2), out_token, mmsg.group(2)) else: yield (0, out_token, line) else: return
class PostgresExplainLexer(RegexLexer): """
Handle PostgreSQL EXPLAIN output """
# This match estimated cost and effectively measured counters with ANALYZE # Then, we move to instrumentation state
(r'(cost)(=?)', bygroups(Name.Class, Punctuation), 'instrumentation'),
(r'(actual)( )(=?)', bygroups(Name.Class, Whitespace, Punctuation), 'instrumentation'),
# We move to sort state in order to emphasize specific keywords (especially disk access)
(r'(Sort Method)(: )', bygroups(Comment.Preproc, Punctuation), 'sort'),
# These keywords can be followed by an object, like a table
(r'(Sort Key|Group Key|Presorted Key|Hash Key)(:)( )',
bygroups(Comment.Preproc, Punctuation, Whitespace), 'object_name'),
(r'(Cache Key|Cache Mode)(:)( )', bygroups(Comment, Punctuation, Whitespace), 'object_name'),
# These keywords can be followed by a predicate
(words(('Join Filter', 'Subplans Removed', 'Filter', 'Merge Cond', 'Hash Cond', 'Index Cond', 'Recheck Cond', 'Heap Blocks', 'TID Cond', 'Run Condition', 'Order By', 'Function Call', 'Table Function Call', 'Inner Unique', 'Params Evaluated', 'Single Copy', 'Sampling', 'One-Time Filter', 'Output', 'Relations', 'Remote SQL'), suffix=r'\b'),
Comment.Preproc, 'predicate'),
# Special keyword to handle ON CONFLICT
(r'Conflict ', Comment.Preproc, 'conflict'),
# Special keyword for InitPlan or SubPlan
(r'(InitPlan|SubPlan)( )(\d+)( )',
bygroups(Keyword, Whitespace, Number.Integer, Whitespace), 'init_plan'),
], 'expression': [ # matches any kind of parenthesized expression # the first opening paren is matched by the 'caller'
(r'\(', Punctuation, '#push'),
(r'\)', Punctuation, '#pop'),
(r'(never executed)', Name.Exception),
(r'[^)(]+', Comment),
], 'object_name': [
# This is a cost or analyze measure
(r'(\(cost)(=?)', bygroups(Name.Class, Punctuation), 'instrumentation'),
(r'(\(actual)( )(=?)', bygroups(Name.Class, Whitespace, Punctuation), 'instrumentation'),
# if object_name is parenthesized, mark opening paren as # punctuation, call 'expression', and exit state
(r'\(', Punctuation, 'expression'),
(r'(on)', Punctuation), # matches possibly schema-qualified table and column names
(r'\w+(\.\w+)*( USING \S+| \w+ USING \S+)', Name.Variable),
(r'\"?\w+\"?(?:\.\"?\w+\"?)?', Name.Variable),
(r'\'\S*\'', Name.Variable),
# if we encounter a comma, another object is listed
(r',\n', Punctuation, 'object_name'),
(r',', Punctuation, 'object_name'),
tokens = { 'root': [
(r'\s+', Whitespace),
(r'--.*?$\n?', Comment.Single),
(r'/\*', Comment.Multiline, 'multiline-comments'),
(words(_tsql_builtins.OPERATORS), Operator),
(words(_tsql_builtins.OPERATOR_WORDS, suffix=r'\b'), Operator.Word),
(words(_tsql_builtins.TYPES, suffix=r'\b'), Name.Class),
(words(_tsql_builtins.FUNCTIONS, suffix=r'\b'), Name.Function),
(r'(goto)(\s+)(\w+\b)', bygroups(Keyword, Whitespace, Name.Label)),
(words(_tsql_builtins.KEYWORDS, suffix=r'\b'), Keyword),
(r'(\[)([^]]+)(\])', bygroups(Operator, Name, Operator)),
(r'0x[0-9a-f]+', Number.Hex), # Float variant 1, for example: 1., 1.e2, 1.2e3
(r'[0-9]+\.[0-9]*(e[+-]?[0-9]+)?', Number.Float), # Float variant 2, for example: .1, .1e2
(r'\.[0-9]+(e[+-]?[0-9]+)?', Number.Float), # Float variant 3, for example: 123e45
(r'[0-9]+e[+-]?[0-9]+', Number.Float),
(r'[0-9]+', Number.Integer),
(r"'(''|[^'])*'", String.Single),
(r'"(""|[^"])*"', String.Symbol),
(r'[;(),.]', Punctuation), # Below we use \w even for the first "real" character because # tokens starting with a digit have already been recognized # as Number above.
(r'@@\w+', Name.Builtin),
(r'@\w+', Name.Variable),
(r'(\w+)(:)', bygroups(Name.Label, Punctuation)),
(r'#?#?\w+', Name), # names for temp tables and anything else
(r'\?', Name.Variable.Magic), # parameter for prepared statements
], 'multiline-comments': [
(r'/\*', Comment.Multiline, 'multiline-comments'),
(r'\*/', Comment.Multiline, '#pop'),
(r'[^/*]+', Comment.Multiline),
(r'[/*]', Comment.Multiline)
]
}
def analyse_text(text):
rating = 0 if tsql_declare_re.search(text): # Found T-SQL variable declaration.
rating = 1.0 else:
name_between_backtick_count = len(
name_between_backtick_re.findall(text))
name_between_bracket_count = len(
name_between_bracket_re.findall(text)) # We need to check if there are any names using # backticks or brackets, as otherwise both are 0 # and 0 >= 2 * 0, so we would always assume it's true
dialect_name_count = name_between_backtick_count + name_between_bracket_count if dialect_name_count >= 1 and \
name_between_bracket_count >= 2 * name_between_backtick_count: # Found at least twice as many [name] as `name`.
rating += 0.5 elif name_between_bracket_count > name_between_backtick_count:
rating += 0.2 elif name_between_bracket_count > 0:
rating += 0.1 if tsql_variable_re.search(text) isnotNone:
rating += 0.1 if tsql_go_re.search(text) isnotNone:
rating += 0.1 return rating
class MySqlLexer(RegexLexer): """The Oracle MySQL lexer.
This lexer does not attempt to maintain strict compatibility with
MariaDB syntax or keywords. Although MySQL and MariaDB's common code
history suggests there may be significant overlap between the two,
compatibility between the two isnot a target for this lexer. """
# Numeric literals
(r'[0-9]+\.[0-9]*(e[+-]?[0-9]+)?', Number.Float), # Mandatory integer, optional fraction and exponent
(r'[0-9]*\.[0-9]+(e[+-]?[0-9]+)?', Number.Float), # Mandatory fraction, optional integer and exponent
(r'[0-9]+e[+-]?[0-9]+', Number.Float), # Exponents with integer significands are still floats
(r'[0-9]+(?=[^0-9a-z$_\u0080-\uffff])', Number.Integer), # Integers that are not in a schema object name
# Date literals
(r"\{\s*d\s*(?P['\"])\s*\d{2}(\d{2})?.?\d{2}.?\d{2}\s*(?P=quote)\s*\}",
Literal.Date),
# Time literals
(r"\{\s*t\s*(?P['\"])\s*(?:\d+\s+)?\d{1,2}.?\d{1,2}.?\d{1,2}(\.\d*)?\s*(?P=quote)\s*\}",
Literal.Date),
# Timestamp literals
(
r"\{\s*ts\s*(?P['\"])\s*"
r"\d{2}(?:\d{2})?.?\d{2}.?\d{2}"# Date part
r"\s+"# Whitespace between date and time
r"\d{1,2}.?\d{1,2}.?\d{1,2}(\.\d*)?"# Time part
r"\s*(?P=quote)\s*\}",
Literal.Date
),
# Exceptions; these words tokenize differently in different contexts.
(r'\b(set)(?!\s*\()', Keyword),
(r'\b(character)(\s+)(set)\b', bygroups(Keyword, Whitespace, Keyword)), # In all other known cases, "SET" is tokenized by MYSQL_DATATYPES.
# Schema object names # # Note: Although the first regex supports unquoted all-numeric # identifiers, this will not be a problem in practice because # numeric literals have already been handled above. #
('[0-9a-z$_\u0080-\uffff]+', Name),
(r'`', Name.Quoted, 'schema-object-name'),
# Schema object name substates # ---------------------------- # # "Name.Quoted" and "Name.Quoted.Escape" are non-standard but # formatters will style them as "Name" by default but add # additional styles based on the token name. This gives users # flexibility to add custom styles as desired. # 'schema-object-name': [
(r'[^`]+', Name.Quoted),
(r'``', Name.Quoted.Escape),
(r'`', Name.Quoted, '#pop'),
],
}
def analyse_text(text):
rating = 0
name_between_backtick_count = len(
name_between_backtick_re.findall(text))
name_between_bracket_count = len(
name_between_bracket_re.findall(text)) # Same logic as above in the TSQL analysis
dialect_name_count = name_between_backtick_count + name_between_bracket_count if dialect_name_count >= 1 and \
name_between_backtick_count >= 2 * name_between_bracket_count: # Found at least twice as many `name` as [name].
rating += 0.5 elif name_between_backtick_count > name_between_bracket_count:
rating += 0.2 elif name_between_backtick_count > 0:
rating += 0.1 return rating
class SqliteConsoleLexer(Lexer): """
Lexer for example sessions using sqlite3. """
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.