Skip to content
Projects
Groups
Snippets
Help
This project
Loading...
Sign in / Register
Toggle navigation
N
news
Project
Overview
Details
Activity
Cycle Analytics
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Charts
Issues
0
Issues
0
List
Board
Labels
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Charts
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Charts
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
Sartika Aritonang
news
Commits
4df54651
Commit
4df54651
authored
May 29, 2020
by
Sartika Aritonang
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
Upload New File
parent
bfc7be00
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
453 additions
and
0 deletions
+453
-0
grouping.py
stbi/Lib/site-packages/sqlparse/engine/grouping.py
+453
-0
No files found.
stbi/Lib/site-packages/sqlparse/engine/grouping.py
0 → 100644
View file @
4df54651
# -*- coding: utf-8 -*-
#
# Copyright (C) 2009-2018 the sqlparse authors and contributors
# <see AUTHORS file>
#
# This module is part of python-sqlparse and is released under
# the BSD License: https://opensource.org/licenses/BSD-3-Clause
from
sqlparse
import
sql
from
sqlparse
import
tokens
as
T
from
sqlparse.utils
import
recurse
,
imt
T_NUMERICAL
=
(
T
.
Number
,
T
.
Number
.
Integer
,
T
.
Number
.
Float
)
T_STRING
=
(
T
.
String
,
T
.
String
.
Single
,
T
.
String
.
Symbol
)
T_NAME
=
(
T
.
Name
,
T
.
Name
.
Placeholder
)
def
_group_matching
(
tlist
,
cls
):
"""Groups Tokens that have beginning and end."""
opens
=
[]
tidx_offset
=
0
for
idx
,
token
in
enumerate
(
list
(
tlist
)):
tidx
=
idx
-
tidx_offset
if
token
.
is_whitespace
:
# ~50% of tokens will be whitespace. Will checking early
# for them avoid 3 comparisons, but then add 1 more comparison
# for the other ~50% of tokens...
continue
if
token
.
is_group
and
not
isinstance
(
token
,
cls
):
# Check inside previously grouped (i.e. parenthesis) if group
# of different type is inside (i.e., case). though ideally should
# should check for all open/close tokens at once to avoid recursion
_group_matching
(
token
,
cls
)
continue
if
token
.
match
(
*
cls
.
M_OPEN
):
opens
.
append
(
tidx
)
elif
token
.
match
(
*
cls
.
M_CLOSE
):
try
:
open_idx
=
opens
.
pop
()
except
IndexError
:
# this indicates invalid sql and unbalanced tokens.
# instead of break, continue in case other "valid" groups exist
continue
close_idx
=
tidx
tlist
.
group_tokens
(
cls
,
open_idx
,
close_idx
)
tidx_offset
+=
close_idx
-
open_idx
def
group_brackets
(
tlist
):
_group_matching
(
tlist
,
sql
.
SquareBrackets
)
def
group_parenthesis
(
tlist
):
_group_matching
(
tlist
,
sql
.
Parenthesis
)
def
group_case
(
tlist
):
_group_matching
(
tlist
,
sql
.
Case
)
def
group_if
(
tlist
):
_group_matching
(
tlist
,
sql
.
If
)
def
group_for
(
tlist
):
_group_matching
(
tlist
,
sql
.
For
)
def
group_begin
(
tlist
):
_group_matching
(
tlist
,
sql
.
Begin
)
def
group_typecasts
(
tlist
):
def
match
(
token
):
return
token
.
match
(
T
.
Punctuation
,
'::'
)
def
valid
(
token
):
return
token
is
not
None
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
return
pidx
,
nidx
valid_prev
=
valid_next
=
valid
_group
(
tlist
,
sql
.
Identifier
,
match
,
valid_prev
,
valid_next
,
post
)
def
group_tzcasts
(
tlist
):
def
match
(
token
):
return
token
.
ttype
==
T
.
Keyword
.
TZCast
def
valid
(
token
):
return
token
is
not
None
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
return
pidx
,
nidx
_group
(
tlist
,
sql
.
Identifier
,
match
,
valid
,
valid
,
post
)
def
group_typed_literal
(
tlist
):
# definitely not complete, see e.g.:
# https://docs.microsoft.com/en-us/sql/odbc/reference/appendixes/interval-literal-syntax
# https://docs.microsoft.com/en-us/sql/odbc/reference/appendixes/interval-literals
# https://www.postgresql.org/docs/9.1/datatype-datetime.html
# https://www.postgresql.org/docs/9.1/functions-datetime.html
def
match
(
token
):
return
imt
(
token
,
m
=
sql
.
TypedLiteral
.
M_OPEN
)
def
match_to_extend
(
token
):
return
isinstance
(
token
,
sql
.
TypedLiteral
)
def
valid_prev
(
token
):
return
token
is
not
None
def
valid_next
(
token
):
return
token
is
not
None
and
token
.
match
(
*
sql
.
TypedLiteral
.
M_CLOSE
)
def
valid_final
(
token
):
return
token
is
not
None
and
token
.
match
(
*
sql
.
TypedLiteral
.
M_EXTEND
)
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
return
tidx
,
nidx
_group
(
tlist
,
sql
.
TypedLiteral
,
match
,
valid_prev
,
valid_next
,
post
,
extend
=
False
)
_group
(
tlist
,
sql
.
TypedLiteral
,
match_to_extend
,
valid_prev
,
valid_final
,
post
,
extend
=
True
)
def
group_period
(
tlist
):
def
match
(
token
):
return
token
.
match
(
T
.
Punctuation
,
'.'
)
def
valid_prev
(
token
):
sqlcls
=
sql
.
SquareBrackets
,
sql
.
Identifier
ttypes
=
T
.
Name
,
T
.
String
.
Symbol
return
imt
(
token
,
i
=
sqlcls
,
t
=
ttypes
)
def
valid_next
(
token
):
# issue261, allow invalid next token
return
True
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
# next_ validation is being performed here. issue261
sqlcls
=
sql
.
SquareBrackets
,
sql
.
Function
ttypes
=
T
.
Name
,
T
.
String
.
Symbol
,
T
.
Wildcard
next_
=
tlist
[
nidx
]
if
nidx
is
not
None
else
None
valid_next
=
imt
(
next_
,
i
=
sqlcls
,
t
=
ttypes
)
return
(
pidx
,
nidx
)
if
valid_next
else
(
pidx
,
tidx
)
_group
(
tlist
,
sql
.
Identifier
,
match
,
valid_prev
,
valid_next
,
post
)
def
group_as
(
tlist
):
def
match
(
token
):
return
token
.
is_keyword
and
token
.
normalized
==
'AS'
def
valid_prev
(
token
):
return
token
.
normalized
==
'NULL'
or
not
token
.
is_keyword
def
valid_next
(
token
):
ttypes
=
T
.
DML
,
T
.
DDL
,
T
.
CTE
return
not
imt
(
token
,
t
=
ttypes
)
and
token
is
not
None
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
return
pidx
,
nidx
_group
(
tlist
,
sql
.
Identifier
,
match
,
valid_prev
,
valid_next
,
post
)
def
group_assignment
(
tlist
):
def
match
(
token
):
return
token
.
match
(
T
.
Assignment
,
':='
)
def
valid
(
token
):
return
token
is
not
None
and
token
.
ttype
not
in
(
T
.
Keyword
)
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
m_semicolon
=
T
.
Punctuation
,
';'
snidx
,
_
=
tlist
.
token_next_by
(
m
=
m_semicolon
,
idx
=
nidx
)
nidx
=
snidx
or
nidx
return
pidx
,
nidx
valid_prev
=
valid_next
=
valid
_group
(
tlist
,
sql
.
Assignment
,
match
,
valid_prev
,
valid_next
,
post
)
def
group_comparison
(
tlist
):
sqlcls
=
(
sql
.
Parenthesis
,
sql
.
Function
,
sql
.
Identifier
,
sql
.
Operation
)
ttypes
=
T_NUMERICAL
+
T_STRING
+
T_NAME
def
match
(
token
):
return
token
.
ttype
==
T
.
Operator
.
Comparison
def
valid
(
token
):
if
imt
(
token
,
t
=
ttypes
,
i
=
sqlcls
):
return
True
elif
token
and
token
.
is_keyword
and
token
.
normalized
==
'NULL'
:
return
True
else
:
return
False
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
return
pidx
,
nidx
valid_prev
=
valid_next
=
valid
_group
(
tlist
,
sql
.
Comparison
,
match
,
valid_prev
,
valid_next
,
post
,
extend
=
False
)
@recurse
(
sql
.
Identifier
)
def
group_identifier
(
tlist
):
ttypes
=
(
T
.
String
.
Symbol
,
T
.
Name
)
tidx
,
token
=
tlist
.
token_next_by
(
t
=
ttypes
)
while
token
:
tlist
.
group_tokens
(
sql
.
Identifier
,
tidx
,
tidx
)
tidx
,
token
=
tlist
.
token_next_by
(
t
=
ttypes
,
idx
=
tidx
)
def
group_arrays
(
tlist
):
sqlcls
=
sql
.
SquareBrackets
,
sql
.
Identifier
,
sql
.
Function
ttypes
=
T
.
Name
,
T
.
String
.
Symbol
def
match
(
token
):
return
isinstance
(
token
,
sql
.
SquareBrackets
)
def
valid_prev
(
token
):
return
imt
(
token
,
i
=
sqlcls
,
t
=
ttypes
)
def
valid_next
(
token
):
return
True
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
return
pidx
,
tidx
_group
(
tlist
,
sql
.
Identifier
,
match
,
valid_prev
,
valid_next
,
post
,
extend
=
True
,
recurse
=
False
)
def
group_operator
(
tlist
):
ttypes
=
T_NUMERICAL
+
T_STRING
+
T_NAME
sqlcls
=
(
sql
.
SquareBrackets
,
sql
.
Parenthesis
,
sql
.
Function
,
sql
.
Identifier
,
sql
.
Operation
,
sql
.
TypedLiteral
)
def
match
(
token
):
return
imt
(
token
,
t
=
(
T
.
Operator
,
T
.
Wildcard
))
def
valid
(
token
):
return
imt
(
token
,
i
=
sqlcls
,
t
=
ttypes
)
\
or
(
token
and
token
.
match
(
T
.
Keyword
,
(
'CURRENT_DATE'
,
'CURRENT_TIME'
,
'CURRENT_TIMESTAMP'
)))
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
tlist
[
tidx
]
.
ttype
=
T
.
Operator
return
pidx
,
nidx
valid_prev
=
valid_next
=
valid
_group
(
tlist
,
sql
.
Operation
,
match
,
valid_prev
,
valid_next
,
post
,
extend
=
False
)
def
group_identifier_list
(
tlist
):
m_role
=
T
.
Keyword
,
(
'null'
,
'role'
)
sqlcls
=
(
sql
.
Function
,
sql
.
Case
,
sql
.
Identifier
,
sql
.
Comparison
,
sql
.
IdentifierList
,
sql
.
Operation
)
ttypes
=
(
T_NUMERICAL
+
T_STRING
+
T_NAME
+
(
T
.
Keyword
,
T
.
Comment
,
T
.
Wildcard
))
def
match
(
token
):
return
token
.
match
(
T
.
Punctuation
,
','
)
def
valid
(
token
):
return
imt
(
token
,
i
=
sqlcls
,
m
=
m_role
,
t
=
ttypes
)
def
post
(
tlist
,
pidx
,
tidx
,
nidx
):
return
pidx
,
nidx
valid_prev
=
valid_next
=
valid
_group
(
tlist
,
sql
.
IdentifierList
,
match
,
valid_prev
,
valid_next
,
post
,
extend
=
True
)
@recurse
(
sql
.
Comment
)
def
group_comments
(
tlist
):
tidx
,
token
=
tlist
.
token_next_by
(
t
=
T
.
Comment
)
while
token
:
eidx
,
end
=
tlist
.
token_not_matching
(
lambda
tk
:
imt
(
tk
,
t
=
T
.
Comment
)
or
tk
.
is_whitespace
,
idx
=
tidx
)
if
end
is
not
None
:
eidx
,
end
=
tlist
.
token_prev
(
eidx
,
skip_ws
=
False
)
tlist
.
group_tokens
(
sql
.
Comment
,
tidx
,
eidx
)
tidx
,
token
=
tlist
.
token_next_by
(
t
=
T
.
Comment
,
idx
=
tidx
)
@recurse
(
sql
.
Where
)
def
group_where
(
tlist
):
tidx
,
token
=
tlist
.
token_next_by
(
m
=
sql
.
Where
.
M_OPEN
)
while
token
:
eidx
,
end
=
tlist
.
token_next_by
(
m
=
sql
.
Where
.
M_CLOSE
,
idx
=
tidx
)
if
end
is
None
:
end
=
tlist
.
_groupable_tokens
[
-
1
]
else
:
end
=
tlist
.
tokens
[
eidx
-
1
]
# TODO: convert this to eidx instead of end token.
# i think above values are len(tlist) and eidx-1
eidx
=
tlist
.
token_index
(
end
)
tlist
.
group_tokens
(
sql
.
Where
,
tidx
,
eidx
)
tidx
,
token
=
tlist
.
token_next_by
(
m
=
sql
.
Where
.
M_OPEN
,
idx
=
tidx
)
@recurse
()
def
group_aliased
(
tlist
):
I_ALIAS
=
(
sql
.
Parenthesis
,
sql
.
Function
,
sql
.
Case
,
sql
.
Identifier
,
sql
.
Operation
,
sql
.
Comparison
)
tidx
,
token
=
tlist
.
token_next_by
(
i
=
I_ALIAS
,
t
=
T
.
Number
)
while
token
:
nidx
,
next_
=
tlist
.
token_next
(
tidx
)
if
isinstance
(
next_
,
sql
.
Identifier
):
tlist
.
group_tokens
(
sql
.
Identifier
,
tidx
,
nidx
,
extend
=
True
)
tidx
,
token
=
tlist
.
token_next_by
(
i
=
I_ALIAS
,
t
=
T
.
Number
,
idx
=
tidx
)
@recurse
(
sql
.
Function
)
def
group_functions
(
tlist
):
has_create
=
False
has_table
=
False
for
tmp_token
in
tlist
.
tokens
:
if
tmp_token
.
value
==
'CREATE'
:
has_create
=
True
if
tmp_token
.
value
==
'TABLE'
:
has_table
=
True
if
has_create
and
has_table
:
return
tidx
,
token
=
tlist
.
token_next_by
(
t
=
T
.
Name
)
while
token
:
nidx
,
next_
=
tlist
.
token_next
(
tidx
)
if
isinstance
(
next_
,
sql
.
Parenthesis
):
tlist
.
group_tokens
(
sql
.
Function
,
tidx
,
nidx
)
tidx
,
token
=
tlist
.
token_next_by
(
t
=
T
.
Name
,
idx
=
tidx
)
def
group_order
(
tlist
):
"""Group together Identifier and Asc/Desc token"""
tidx
,
token
=
tlist
.
token_next_by
(
t
=
T
.
Keyword
.
Order
)
while
token
:
pidx
,
prev_
=
tlist
.
token_prev
(
tidx
)
if
imt
(
prev_
,
i
=
sql
.
Identifier
,
t
=
T
.
Number
):
tlist
.
group_tokens
(
sql
.
Identifier
,
pidx
,
tidx
)
tidx
=
pidx
tidx
,
token
=
tlist
.
token_next_by
(
t
=
T
.
Keyword
.
Order
,
idx
=
tidx
)
@recurse
()
def
align_comments
(
tlist
):
tidx
,
token
=
tlist
.
token_next_by
(
i
=
sql
.
Comment
)
while
token
:
pidx
,
prev_
=
tlist
.
token_prev
(
tidx
)
if
isinstance
(
prev_
,
sql
.
TokenList
):
tlist
.
group_tokens
(
sql
.
TokenList
,
pidx
,
tidx
,
extend
=
True
)
tidx
=
pidx
tidx
,
token
=
tlist
.
token_next_by
(
i
=
sql
.
Comment
,
idx
=
tidx
)
def
group_values
(
tlist
):
tidx
,
token
=
tlist
.
token_next_by
(
m
=
(
T
.
Keyword
,
'VALUES'
))
start_idx
=
tidx
end_idx
=
-
1
while
token
:
if
isinstance
(
token
,
sql
.
Parenthesis
):
end_idx
=
tidx
tidx
,
token
=
tlist
.
token_next
(
tidx
)
if
end_idx
!=
-
1
:
tlist
.
group_tokens
(
sql
.
Values
,
start_idx
,
end_idx
,
extend
=
True
)
def
group
(
stmt
):
for
func
in
[
group_comments
,
# _group_matching
group_brackets
,
group_parenthesis
,
group_case
,
group_if
,
group_for
,
group_begin
,
group_functions
,
group_where
,
group_period
,
group_arrays
,
group_identifier
,
group_order
,
group_typecasts
,
group_tzcasts
,
group_typed_literal
,
group_operator
,
group_comparison
,
group_as
,
group_aliased
,
group_assignment
,
align_comments
,
group_identifier_list
,
group_values
,
]:
func
(
stmt
)
return
stmt
def
_group
(
tlist
,
cls
,
match
,
valid_prev
=
lambda
t
:
True
,
valid_next
=
lambda
t
:
True
,
post
=
None
,
extend
=
True
,
recurse
=
True
):
"""Groups together tokens that are joined by a middle token. i.e. x < y"""
tidx_offset
=
0
pidx
,
prev_
=
None
,
None
for
idx
,
token
in
enumerate
(
list
(
tlist
)):
tidx
=
idx
-
tidx_offset
if
token
.
is_whitespace
:
continue
if
recurse
and
token
.
is_group
and
not
isinstance
(
token
,
cls
):
_group
(
token
,
cls
,
match
,
valid_prev
,
valid_next
,
post
,
extend
)
if
match
(
token
):
nidx
,
next_
=
tlist
.
token_next
(
tidx
)
if
prev_
and
valid_prev
(
prev_
)
and
valid_next
(
next_
):
from_idx
,
to_idx
=
post
(
tlist
,
pidx
,
tidx
,
nidx
)
grp
=
tlist
.
group_tokens
(
cls
,
from_idx
,
to_idx
,
extend
=
extend
)
tidx_offset
+=
to_idx
-
from_idx
pidx
,
prev_
=
from_idx
,
grp
continue
pidx
,
prev_
=
tidx
,
token
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment