-
Notifications
You must be signed in to change notification settings - Fork 590
/
__init__.py
174 lines (128 loc) · 4.58 KB
/
__init__.py
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
from __future__ import annotations
import functools
import json
from typing import TYPE_CHECKING, Optional
import ibis
from ibis.common.grounds import Concrete
try:
import importlib_resources as resources
except ImportError:
from importlib import resources
if TYPE_CHECKING:
import ibis.expr.types as ir
from ibis.backends import BaseBackend
# These backends load the data directly using `read_csv`/`read_parquet`. All
# other backends load the data using pyarrow, then passing it off to
# `create_table`.
_DIRECT_BACKENDS = frozenset({"duckdb", "polars"})
class Example(Concrete):
name: str
help: Optional[str]
def fetch(
self,
*,
table_name: str | None = None,
backend: BaseBackend | None = None,
) -> ir.Table:
if backend is None:
backend = ibis.get_backend()
name = self.name
if table_name is None:
table_name = name
board = _get_board()
(path,) = board.pin_download(name)
if backend.name in _DIRECT_BACKENDS:
# Read directly into these backends. This helps reduce memory
# usage, making the larger example datasets easier to work with.
if path.endswith(".parquet"):
return backend.read_parquet(path, table_name=table_name)
else:
return backend.read_csv(path, table_name=table_name)
else:
import pyarrow_hotfix # noqa: F401
if path.endswith(".parquet"):
import pyarrow.parquet
table = pyarrow.parquet.read_table(path)
else:
import pyarrow.csv
# The convert options lets pyarrow treat empty strings as null for
# string columns, but not quoted empty strings.
table = pyarrow.csv.read_csv(
path,
convert_options=pyarrow.csv.ConvertOptions(
strings_can_be_null=True,
quoted_strings_can_be_null=False,
),
)
# All null columns are inferred as null-type, but not all
# backends support null-type columns. Cast to an all-null
# string column instead.
for i, field in enumerate(table.schema):
if pyarrow.types.is_null(field.type):
table = table.set_column(i, field.name, table[i].cast("string"))
# TODO: It should be possible to avoid this memtable call, once all
# backends support passing a `pyarrow.Table` to `create_table`
# directly.
obj = ibis.memtable(table)
return backend.create_table(table_name, obj, temp=True, overwrite=True)
_FETCH_DOCSTRING_TEMPLATE = """\
Fetch the {name} example.
Parameters
----------
table_name
The table name to use, defaults to a generated table name.
backend
The backend to load the example into. Defaults to the default backend.
Returns
-------
ir.Table
Table expression
Examples
--------
>>> import ibis
>>> t = ibis.examples.{name}.fetch()
"""
_BUCKET = "ibis-pins"
@functools.cache
def _get_metadata():
return json.loads(resources.files(__name__).joinpath("metadata.json").read_text())
@functools.cache
def _get_board():
import pins
return pins.board(
"gcs", _BUCKET, storage_options={"cache_timeout": 0, "token": "anon"}
)
@functools.cache
def __dir__() -> list[str]:
return sorted(_get_metadata().keys())
class Zones(Concrete):
name: str
help: Optional[str]
def fetch(
self,
*,
table_name: str | None = None,
backend: BaseBackend | None = None,
) -> ir.Table:
if backend is None:
backend = ibis.get_backend()
name = self.name
if table_name is None:
table_name = name
board = _get_board()
(path,) = board.pin_download(name)
return backend.read_geo(path)
zones = Zones("zones", help="Taxi zones in New York City (EPSG:2263)")
def __getattr__(name: str) -> Example:
try:
meta = _get_metadata()
description = meta[name].get("description")
fields = {"__doc__": description} if description is not None else {}
example_class = type(name, (Example,), fields)
example_class.fetch.__doc__ = _FETCH_DOCSTRING_TEMPLATE.format(name=name)
example = example_class(name=name, help=description)
setattr(ibis.examples, name, example)
except Exception as e: # noqa: BLE001
raise AttributeError(name) from e
else:
return example