Coverage for python/lsst/daf/butler/core/ddl.py : 47%

Hot-keys on this page
r m x p toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
1# This file is part of daf_butler.
2#
3# Developed for the LSST Data Management System.
4# This product includes software developed by the LSST Project
5# (http://www.lsst.org).
6# See the COPYRIGHT file at the top-level directory of this distribution
7# for details of code ownership.
8#
9# This program is free software: you can redistribute it and/or modify
10# it under the terms of the GNU General Public License as published by
11# the Free Software Foundation, either version 3 of the License, or
12# (at your option) any later version.
13#
14# This program is distributed in the hope that it will be useful,
15# but WITHOUT ANY WARRANTY; without even the implied warranty of
16# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17# GNU General Public License for more details.
18#
19# You should have received a copy of the GNU General Public License
20# along with this program. If not, see <http://www.gnu.org/licenses/>.
21"""Classes for representing SQL data-definition language (DDL; "CREATE TABLE",
22etc.) in Python.
24This provides an extra layer on top of SQLAlchemy's classes for these concepts,
25because we need a level of indirection between logical tables and the actual
26SQL, and SQLAlchemy's DDL classes always map 1-1 to SQL.
28We've opted for the rather more obscure "ddl" as the name of this module
29instead of "schema" because the latter is too overloaded; in most SQL
30databases, a "schema" is also another term for a namespace.
31"""
32from __future__ import annotations
34__all__ = ("TableSpec", "FieldSpec", "ForeignKeySpec", "Base64Bytes", "Base64Region",
35 "AstropyTimeNsecTai")
37from base64 import b64encode, b64decode
38import logging
39from math import ceil
40from dataclasses import dataclass
41from typing import Any, Callable, Iterable, List, Optional, Set, Tuple, Type, TYPE_CHECKING, Union
43import sqlalchemy
44import astropy.time
46from lsst.sphgeom import ConvexPolygon
47from .config import Config
48from .exceptions import ValidationError
49from . import time_utils
50from .utils import iterable, stripIfNotNone
51from .named import NamedValueSet
53if TYPE_CHECKING: 53 ↛ 54line 53 didn't jump to line 54, because the condition on line 53 was never true
54 from .timespan import DatabaseTimespanRepresentation
57_LOG = logging.getLogger(__name__)
60class SchemaValidationError(ValidationError):
61 """Exceptions used to indicate problems in Registry schema configuration.
62 """
64 @classmethod
65 def translate(cls, caught: Type[Exception], message: str) -> Callable:
66 """A decorator that re-raises exceptions as `SchemaValidationError`.
68 Decorated functions must be class or instance methods, with a
69 ``config`` parameter as their first argument. This will be passed
70 to ``message.format()`` as a keyword argument, along with ``err``,
71 the original exception.
73 Parameters
74 ----------
75 caught : `type` (`Exception` subclass)
76 The type of exception to catch.
77 message : `str`
78 A `str.format` string that may contain named placeholders for
79 ``config``, ``err``, or any keyword-only argument accepted by
80 the decorated function.
81 """
82 def decorate(func: Callable) -> Callable:
83 def decorated(self: Any, config: Config, *args: Any, **kwds: Any) -> Any:
84 try:
85 return func(self, config, *args, **kwds)
86 except caught as err:
87 raise cls(message.format(config=str(config), err=err))
88 return decorated
89 return decorate
92class Base64Bytes(sqlalchemy.TypeDecorator):
93 """A SQLAlchemy custom type that maps Python `bytes` to a base64-encoded
94 `sqlalchemy.Text` field.
95 """
97 impl = sqlalchemy.Text
99 def __init__(self, nbytes: int, *args: Any, **kwargs: Any):
100 length = 4*ceil(nbytes/3) if self.impl == sqlalchemy.String else None
101 super().__init__(*args, length=length, **kwargs)
102 self.nbytes = nbytes
104 def process_bind_param(self, value: Optional[bytes], dialect: sqlalchemy.engine.Dialect
105 ) -> Optional[str]:
106 # 'value' is native `bytes`. We want to encode that to base64 `bytes`
107 # and then ASCII `str`, because `str` is what SQLAlchemy expects for
108 # String fields.
109 if value is None:
110 return None
111 if not isinstance(value, bytes):
112 raise TypeError(
113 f"Base64Bytes fields require 'bytes' values; got '{value}' with type {type(value)}."
114 )
115 return b64encode(value).decode("ascii")
117 def process_result_value(self, value: Optional[str], dialect: sqlalchemy.engine.Dialect
118 ) -> Optional[bytes]:
119 # 'value' is a `str` that must be ASCII because it's base64-encoded.
120 # We want to transform that to base64-encoded `bytes` and then
121 # native `bytes`.
122 return b64decode(value.encode("ascii")) if value is not None else None
125class Base64Region(Base64Bytes):
126 """A SQLAlchemy custom type that maps Python `sphgeom.ConvexPolygon` to a
127 base64-encoded `sqlalchemy.String`.
128 """
130 def process_bind_param(self, value: Optional[ConvexPolygon], dialect: sqlalchemy.engine.Dialect
131 ) -> Optional[str]:
132 if value is None:
133 return None
134 return super().process_bind_param(value.encode(), dialect)
136 def process_result_value(self, value: Optional[str], dialect: sqlalchemy.engine.Dialect
137 ) -> Optional[ConvexPolygon]:
138 if value is None:
139 return None
140 return ConvexPolygon.decode(super().process_result_value(value, dialect))
143class AstropyTimeNsecTai(sqlalchemy.TypeDecorator):
144 """A SQLAlchemy custom type that maps Python `astropy.time.Time` to a
145 number of nanoseconds since Unix epoch in TAI scale.
146 """
148 impl = sqlalchemy.BigInteger
150 def process_bind_param(self, value: Optional[astropy.time.Time], dialect: sqlalchemy.engine.Dialect
151 ) -> Optional[int]:
152 if value is None:
153 return None
154 if not isinstance(value, astropy.time.Time):
155 raise TypeError(f"Unsupported type: {type(value)}, expected astropy.time.Time")
156 value = time_utils.astropy_to_nsec(value)
157 return value
159 def process_result_value(self, value: Optional[int], dialect: sqlalchemy.engine.Dialect
160 ) -> Optional[astropy.time.Time]:
161 # value is nanoseconds since epoch, or None
162 if value is None:
163 return None
164 value = time_utils.nsec_to_astropy(value)
165 return value
168VALID_CONFIG_COLUMN_TYPES = {
169 "string": sqlalchemy.String,
170 "int": sqlalchemy.BigInteger,
171 "float": sqlalchemy.Float,
172 "region": Base64Region,
173 "bool": sqlalchemy.Boolean,
174 "blob": sqlalchemy.LargeBinary,
175 "datetime": AstropyTimeNsecTai,
176 "hash": Base64Bytes
177}
180@dataclass
181class FieldSpec:
182 """A struct-like class used to define a column in a logical `Registry`
183 table.
184 """
186 name: str
187 """Name of the column."""
189 dtype: type
190 """Type of the column; usually a `type` subclass provided by SQLAlchemy
191 that defines both a Python type and a corresponding precise SQL type.
192 """
194 length: Optional[int] = None
195 """Length of the type in the database, for variable-length types."""
197 nbytes: Optional[int] = None
198 """Natural length used for hash and encoded-region columns, to be converted
199 into the post-encoding length.
200 """
202 primaryKey: bool = False
203 """Whether this field is (part of) its table's primary key."""
205 autoincrement: bool = False
206 """Whether the database should insert automatically incremented values when
207 no value is provided in an INSERT.
208 """
210 nullable: bool = True
211 """Whether this field is allowed to be NULL."""
213 default: Any = None
214 """A server-side default value for this field.
216 This is passed directly as the ``server_default`` argument to
217 `sqlalchemy.schema.Column`. It does _not_ go through SQLAlchemy's usual
218 type conversion or quoting for Python literals, and should hence be used
219 with care. See the SQLAlchemy documentation for more information.
220 """
222 doc: Optional[str] = None
223 """Documentation for this field."""
225 def __eq__(self, other: Any) -> Union[bool, NotImplemented]:
226 if isinstance(other, FieldSpec):
227 return self.name == other.name
228 else:
229 return NotImplemented
231 def __hash__(self) -> int:
232 return hash(self.name)
234 @classmethod
235 @SchemaValidationError.translate(KeyError, "Missing key {err} in column config '{config}'.")
236 def fromConfig(cls, config: Config, **kwds: Any) -> FieldSpec:
237 """Create a `FieldSpec` from a subset of a `SchemaConfig`.
239 Parameters
240 ----------
241 config: `Config`
242 Configuration describing the column. Nested configuration keys
243 correspond to `FieldSpec` attributes.
244 kwds
245 Additional keyword arguments that provide defaults for values
246 not present in config.
248 Returns
249 -------
250 spec: `FieldSpec`
251 Specification structure for the column.
253 Raises
254 ------
255 SchemaValidationError
256 Raised if configuration keys are missing or have invalid values.
257 """
258 dtype = VALID_CONFIG_COLUMN_TYPES.get(config["type"])
259 if dtype is None:
260 raise SchemaValidationError(f"Invalid field type string: '{config['type']}'.")
261 if not config["name"].islower():
262 raise SchemaValidationError(f"Column name '{config['name']}' is not all lowercase.")
263 self = cls(name=config["name"], dtype=dtype, **kwds)
264 self.length = config.get("length", self.length)
265 self.nbytes = config.get("nbytes", self.nbytes)
266 if self.length is not None and self.nbytes is not None:
267 raise SchemaValidationError(f"Both length and nbytes provided for field '{self.name}'.")
268 self.primaryKey = config.get("primaryKey", self.primaryKey)
269 self.autoincrement = config.get("autoincrement", self.autoincrement)
270 self.nullable = config.get("nullable", False if self.primaryKey else self.nullable)
271 self.doc = stripIfNotNone(config.get("doc", None))
272 return self
274 def isStringType(self) -> bool:
275 """Indicate that this is a sqlalchemy.String field spec.
277 Returns
278 -------
279 isString : `bool`
280 The field refers to a `sqlalchemy.String` and not any other type.
281 This can return `False` even if the object was created with a
282 string type if it has been decided that it should be implemented
283 as a `sqlalchemy.Text` type.
284 """
285 if self.dtype == sqlalchemy.String:
286 # For short strings retain them as strings
287 if self.dtype == sqlalchemy.String and self.length and self.length <= 32:
288 return True
289 return False
291 def getSizedColumnType(self) -> sqlalchemy.types.TypeEngine:
292 """Return a sized version of the column type, utilizing either (or
293 neither) of ``self.length`` and ``self.nbytes``.
295 Returns
296 -------
297 dtype : `sqlalchemy.types.TypeEngine`
298 A SQLAlchemy column type object.
299 """
300 if self.length is not None:
301 # Last chance check that we are only looking at possible String
302 if self.dtype == sqlalchemy.String and not self.isStringType():
303 return sqlalchemy.Text
304 return self.dtype(length=self.length)
305 if self.nbytes is not None:
306 return self.dtype(nbytes=self.nbytes)
307 return self.dtype
310@dataclass
311class ForeignKeySpec:
312 """A struct-like class used to define a foreign key constraint in a logical
313 `Registry` table.
314 """
316 table: str
317 """Name of the target table."""
319 source: Tuple[str, ...]
320 """Tuple of source table column names."""
322 target: Tuple[str, ...]
323 """Tuple of target table column names."""
325 onDelete: Optional[str] = None
326 """SQL clause indicating how to handle deletes to the target table.
328 If not `None` (which indicates that a constraint violation exception should
329 be raised), should be either "SET NULL" or "CASCADE".
330 """
332 addIndex: bool = True
333 """If `True`, create an index on the columns of this foreign key in the
334 source table.
335 """
337 @classmethod
338 @SchemaValidationError.translate(KeyError, "Missing key {err} in foreignKey config '{config}'.")
339 def fromConfig(cls, config: Config) -> ForeignKeySpec:
340 """Create a `ForeignKeySpec` from a subset of a `SchemaConfig`.
342 Parameters
343 ----------
344 config: `Config`
345 Configuration describing the constraint. Nested configuration keys
346 correspond to `ForeignKeySpec` attributes.
348 Returns
349 -------
350 spec: `ForeignKeySpec`
351 Specification structure for the constraint.
353 Raises
354 ------
355 SchemaValidationError
356 Raised if configuration keys are missing or have invalid values.
357 """
358 return cls(table=config["table"],
359 source=tuple(iterable(config["source"])),
360 target=tuple(iterable(config["target"])),
361 onDelete=config.get("onDelete", None))
364@dataclass
365class TableSpec:
366 """A struct-like class used to define a table or table-like
367 query interface.
369 Parameters
370 ----------
371 fields : `Iterable` [ `FieldSpec` ]
372 Specifications for the columns in this table.
373 unique : `Iterable` [ `tuple` [ `str` ] ], optional
374 Non-primary-key unique constraints for the table.
375 indexes: `Iterable` [ `tuple` [ `str` ] ], optional
376 Indexes for the table.
377 foreignKeys : `Iterable` [ `ForeignKeySpec` ], optional
378 Foreign key constraints for the table.
379 exclusion : `Iterable` [ `tuple` [ `str` or `type` ] ]
380 Special constraints that prohibit overlaps between timespans over rows
381 where other columns are equal. These take the same form as unique
382 constraints, but each tuple may contain a single
383 `DatabaseTimespanRepresentation` subclass representing a timespan
384 column.
385 recycleIds : bool, optional
386 If `True`, allow databases that might normally recycle autoincrement
387 IDs to do so (usually better for performance) on any autoincrement
388 field in this table.
389 doc : str, optional
390 Documentation for the table.
391 """
392 def __init__(
393 self, fields: Iterable[FieldSpec], *,
394 unique: Iterable[Tuple[str, ...]] = (),
395 indexes: Iterable[Tuple[str, ...]] = (),
396 foreignKeys: Iterable[ForeignKeySpec] = (),
397 exclusion: Iterable[Tuple[Union[str, Type[DatabaseTimespanRepresentation]], ...]] = (),
398 recycleIds: bool = True,
399 doc: Optional[str] = None,
400 ):
401 self.fields = NamedValueSet(fields)
402 self.unique = set(unique)
403 self.indexes = set(indexes)
404 self.foreignKeys = list(foreignKeys)
405 self.exclusion = set(exclusion)
406 self.recycleIds = recycleIds
407 self.doc = doc
409 fields: NamedValueSet[FieldSpec]
410 """Specifications for the columns in this table."""
412 unique: Set[Tuple[str, ...]]
413 """Non-primary-key unique constraints for the table."""
415 indexes: Set[Tuple[str, ...]]
416 """Indexes for the table."""
418 foreignKeys: List[ForeignKeySpec]
419 """Foreign key constraints for the table."""
421 exclusion: Set[Tuple[Union[str, Type[DatabaseTimespanRepresentation]], ...]]
422 """Exclusion constraints for the table.
424 Exclusion constraints behave mostly like unique constraints, but may
425 contain a database-native Timespan column that is restricted to not overlap
426 across rows (for identical combinations of any non-Timespan columns in the
427 constraint).
428 """
430 recycleIds: bool = True
431 """If `True`, allow databases that might normally recycle autoincrement IDs
432 to do so (usually better for performance) on any autoincrement field in
433 this table.
434 """
436 doc: Optional[str] = None
437 """Documentation for the table."""
439 @classmethod
440 @SchemaValidationError.translate(KeyError, "Missing key {err} in table config '{config}'.")
441 def fromConfig(cls, config: Config) -> TableSpec:
442 """Create a `ForeignKeySpec` from a subset of a `SchemaConfig`.
444 Parameters
445 ----------
446 config: `Config`
447 Configuration describing the constraint. Nested configuration keys
448 correspond to `TableSpec` attributes.
450 Returns
451 -------
452 spec: `TableSpec`
453 Specification structure for the table.
455 Raises
456 ------
457 SchemaValidationError
458 Raised if configuration keys are missing or have invalid values.
459 """
460 return cls(
461 fields=NamedValueSet(FieldSpec.fromConfig(c) for c in config["columns"]),
462 unique={tuple(u) for u in config.get("unique", ())},
463 foreignKeys=[ForeignKeySpec.fromConfig(c) for c in config.get("foreignKeys", ())],
464 doc=stripIfNotNone(config.get("doc")),
465 )