forked from tobymao/sqlglot
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathpython.py
More file actions
409 lines (337 loc) · 14.4 KB
/
Copy pathpython.py
File metadata and controls
409 lines (337 loc) · 14.4 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
import collections
import itertools
import math
from sqlglot import exp, planner, tokens
from sqlglot.dialects.dialect import Dialect
from sqlglot.errors import ExecuteError
from sqlglot.executor.context import Context
from sqlglot.executor.env import ENV
from sqlglot.executor.table import RowReader, Table
from sqlglot.generators.python import PythonGenerator
class PythonExecutor:
def __init__(self, env=None, tables=None):
self.generator = Python().generator(identify=True, comments=False)
self.env = {**ENV, **(env or {})}
self.tables = tables or {}
def execute(self, plan):
finished = set()
queue = set(plan.leaves)
contexts = {}
while queue:
node = queue.pop()
try:
context = self.context(
{
name: table
for dep in node.dependencies
for name, table in contexts[dep].tables.items()
}
)
if isinstance(node, planner.Scan):
contexts[node] = self.scan(node, context)
elif isinstance(node, planner.Aggregate):
contexts[node] = self.aggregate(node, context)
elif isinstance(node, planner.Join):
contexts[node] = self.join(node, context)
elif isinstance(node, planner.Sort):
contexts[node] = self.sort(node, context)
elif isinstance(node, planner.SetOperation):
contexts[node] = self.set_operation(node, context)
else:
raise NotImplementedError
finished.add(node)
for dep in node.dependents:
if all(d in contexts for d in dep.dependencies):
queue.add(dep)
for dep in node.dependencies:
if all(d in finished for d in dep.dependents):
contexts.pop(dep)
except Exception as e:
raise ExecuteError(f"Step '{node.id}' failed: {e}") from e
root = plan.root
return contexts[root].tables[root.name]
def generate(self, expression):
"""Convert a SQL expression into literal Python code and compile it into bytecode."""
if not expression:
return None
sql = self.generator.generate(expression)
return compile(sql, sql, "eval", optimize=2)
def generate_tuple(self, expressions):
"""Convert an array of SQL expressions into tuple of Python byte code."""
if not expressions:
return tuple()
return tuple(self.generate(expression) for expression in expressions)
def context(self, tables):
return Context(tables, env=self.env)
def table(self, expressions):
return Table(
expression.alias_or_name if isinstance(expression, exp.Expr) else expression
for expression in expressions
)
def scan(self, step, context):
source = step.source
if source and isinstance(source, exp.Expr):
source = source.name or source.alias
if source is None:
context, table_iter = self.static()
elif source in context:
if not step.projections and not step.condition:
return self.context({step.name: context.tables[source]})
table_iter = context.table_iter(source)
else:
context, table_iter = self.scan_table(step)
return self.context({step.name: self._project_and_filter(context, step, table_iter)})
def _project_and_filter(self, context, step, table_iter):
sink = self.table(step.projections if step.projections else context.columns)
condition = self.generate(step.condition)
projections = self.generate_tuple(step.projections)
for reader in table_iter:
if len(sink) >= step.limit:
break
if condition and not context.eval(condition):
continue
if projections:
sink.append(context.eval_tuple(projections))
else:
sink.append(reader.row)
return sink
def static(self):
return self.context({}), [RowReader(())]
def scan_table(self, step):
table = self.tables.find(step.source)
context = self.context({step.source.alias_or_name: table})
return context, iter(table)
def join(self, step, context):
source = step.source_name
source_table = context.tables[source]
source_context = self.context({source: source_table})
column_ranges = {source: range(0, len(source_table.columns))}
for name, join in step.joins.items():
table = context.tables[name]
start = max(r.stop for r in column_ranges.values())
column_ranges[name] = range(start, len(table.columns) + start)
join_context = self.context({name: table})
condition = self.generate(join["condition"])
condition_context = (
self.context(
{
name: Table(
source_context.columns + join_context.columns,
column_range=column_range,
)
for name, column_range in column_ranges.items()
}
)
if condition
else None
)
if join.get("source_key"):
table = self.hash_join(
join, source_context, join_context, condition, condition_context
)
else:
table = self.nested_loop_join(
join, source_context, join_context, condition, condition_context
)
source_context = self.context(
{
name: Table(table.columns, table.rows, column_range)
for name, column_range in column_ranges.items()
}
)
if not step.condition and not step.projections:
return source_context
sink = self._project_and_filter(
source_context,
step,
(reader for reader, _ in iter(source_context)),
)
if step.projections:
return self.context({step.name: sink})
else:
return self.context(
{
name: Table(table.columns, sink.rows, table.column_range)
for name, table in source_context.tables.items()
}
)
@staticmethod
def _join_matches(row, condition, condition_context):
if not condition:
return True
condition_context.set_row(row)
return condition_context.eval(condition) is True
def nested_loop_join(self, join, source_context, join_context, condition, condition_context):
table = Table(source_context.columns + join_context.columns)
source_rows = source_context.table.rows
join_rows = join_context.table.rows
matched_source = set()
matched_join = set()
for source_index, source_row in enumerate(source_rows):
for join_index, join_row in enumerate(join_rows):
row = source_row + join_row
if self._join_matches(row, condition, condition_context):
table.append(row)
matched_source.add(source_index)
matched_join.add(join_index)
self._append_unmatched_join_rows(
table, join, source_rows, join_rows, matched_source, matched_join
)
return table
def hash_join(self, join, source_context, join_context, condition, condition_context):
source_key = self.generate_tuple(join["source_key"])
join_key = self.generate_tuple(join["join_key"])
results = collections.defaultdict(lambda: ([], []))
for index, (reader, ctx) in enumerate(source_context):
key = ctx.eval_tuple(source_key)
if all(value is not None for value in key):
results[key][0].append((index, reader.row))
for index, (reader, ctx) in enumerate(join_context):
key = ctx.eval_tuple(join_key)
if all(value is not None for value in key):
results[key][1].append((index, reader.row))
table = Table(source_context.columns + join_context.columns)
matched_source = set()
matched_join = set()
for source_group, join_group in results.values():
for (source_index, source_row), (join_index, join_row) in itertools.product(
source_group, join_group
):
row = source_row + join_row
if self._join_matches(row, condition, condition_context):
table.append(row)
matched_source.add(source_index)
matched_join.add(join_index)
self._append_unmatched_join_rows(
table,
join,
source_context.table.rows,
join_context.table.rows,
matched_source,
matched_join,
)
return table
@staticmethod
def _append_unmatched_join_rows(
table, join, source_rows, join_rows, matched_source, matched_join
):
side = join.get("side")
if side in ("LEFT", "FULL"):
join_nulls = (None,) * (len(table.columns) - len(source_rows[0]) if source_rows else 0)
for index, row in enumerate(source_rows):
if index not in matched_source:
table.append(row + join_nulls)
if side in ("RIGHT", "FULL"):
source_width = len(table.columns) - (len(join_rows[0]) if join_rows else 0)
source_nulls = (None,) * source_width
for index, row in enumerate(join_rows):
if index not in matched_join:
table.append(source_nulls + row)
def aggregate(self, step, context):
group_by = self.generate_tuple(step.group.values())
aggregations = self.generate_tuple(step.aggregations)
operands = self.generate_tuple(step.operands)
if operands:
operand_table = Table(self.table(step.operands).columns)
for reader, ctx in context:
operand_table.append(ctx.eval_tuple(operands))
for i, (a, b) in enumerate(zip(context.table.rows, operand_table.rows)):
context.table.rows[i] = a + b
width = len(context.columns)
context.add_columns(*operand_table.columns)
operand_table = Table(
context.columns,
context.table.rows,
range(width, width + len(operand_table.columns)),
)
context = self.context(
{
None: operand_table,
**context.tables,
}
)
context.sort(group_by)
group = None
start = 0
end = 1
length = len(context.table)
table = self.table(list(step.group) + step.aggregations)
def add_row():
table.append(group + context.eval_tuple(aggregations))
if length:
for i in range(length):
context.set_index(i)
key = context.eval_tuple(group_by)
group = key if group is None else group
end += 1
if key != group:
context.set_range(start, end - 2)
add_row()
group = key
start = end - 2
if len(table.rows) >= step.limit:
break
if i == length - 1:
context.set_range(start, end - 1)
add_row()
elif step.limit > 0 and not group_by:
context.set_range(0, 0)
table.append(context.eval_tuple(aggregations))
context = self.context({step.name: table, **{name: table for name in context.tables}})
if step.projections or step.condition:
return self.scan(step, context)
return context
def sort(self, step, context):
projections = self.generate_tuple(step.projections)
projection_columns = [p.alias_or_name for p in step.projections]
all_columns = list(context.columns) + projection_columns
sink = self.table(all_columns)
for reader, ctx in context:
sink.append(reader.row + ctx.eval_tuple(projections))
sort_ctx = self.context(
{
None: sink,
**{table: sink for table in context.tables},
}
)
sort_ctx.sort(self.generate_tuple(step.key))
if not math.isinf(step.limit):
sort_ctx.table.rows = sort_ctx.table.rows[0 : step.limit]
output = Table(
projection_columns,
rows=[r[len(context.columns) : len(all_columns)] for r in sort_ctx.table.rows],
)
return self.context({step.name: output})
def set_operation(self, step, context):
left = context.tables[step.left]
right = context.tables[step.right]
sink = self.table(left.columns)
if issubclass(step.op, exp.Intersect):
right_counts = collections.Counter(right.rows)
seen = set()
for row in left.rows:
if right_counts[row] and (not step.distinct or row not in seen):
sink.append(row)
seen.add(row)
if not step.distinct:
right_counts[row] -= 1
elif issubclass(step.op, exp.Except):
right_counts = collections.Counter(right.rows)
seen = set()
for row in left.rows:
if right_counts[row] and not step.distinct:
right_counts[row] -= 1
elif not right_counts[row] and (not step.distinct or row not in seen):
sink.append(row)
seen.add(row)
elif issubclass(step.op, exp.Union) and step.distinct:
sink.rows = list(set(left.rows).union(set(right.rows)))
else:
sink.rows = left.rows + right.rows
if not math.isinf(step.limit):
sink.rows = sink.rows[0 : step.limit]
return self.context({step.name: sink})
class Python(Dialect):
class Tokenizer(tokens.Tokenizer):
STRING_ESCAPES = ["\\"]
Generator = PythonGenerator