Skip to content

Commit af0ecb0

Browse files
plamuttswast
plamut
and
tswast
authored
docs: Add migration guide from version 2.x to 3.x (#1027)
* docs: add migration guide from version 2.x. to 3.x * Add a section on typee annotations * Explain additional requirement of pandas extra * Mention new default type for TZ-aware datetimes * rearrange and add a section * start documenting model properties that have changed * add table of changes for pandas and Model Co-authored-by: Tim Swast <[email protected]>
1 parent 9d256d5 commit af0ecb0

File tree

2 files changed

+183
-2
lines changed

2 files changed

+183
-2
lines changed

‎UPGRADING.md‎

Lines changed: 181 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -13,8 +13,188 @@ limitations under the License.
1313

1414
# 3.0.0 Migration Guide
1515

16-
TODO
16+
## New Required Dependencies
1717

18+
Some of the previously optional dependencies are now *required* in `3.x` versions of the
19+
library, namely
20+
[google-cloud-bigquery-storage](https://pypi.org/project/google-cloud-bigquery-storage/)
21+
(minimum version `2.0.0`) and [pyarrow](https://pypi.org/project/pyarrow/) (minimum
22+
version `3.0.0`).
23+
24+
The behavior of some of the package "extras" has thus also changed:
25+
* The `pandas` extra now requires the [db-types](https://pypi.org/project/db-dtypes/)
26+
package.
27+
* The `bqstorage` extra has been preserved for comaptibility reasons, but it is now a
28+
no-op and should be omitted when installing the BigQuery client library.
29+
30+
**Before:**
31+
```
32+
$ pip install google-cloud-bigquery[bqstorage]
33+
```
34+
35+
**After:**
36+
```
37+
$ pip install google-cloud-bigquery
38+
```
39+
40+
* The `bignumeric_type` extra has been removed, as `BIGNUMERIC` type is now
41+
automatically supported. That extra should thus not be used.
42+
43+
**Before:**
44+
```
45+
$ pip install google-cloud-bigquery[bignumeric_type]
46+
```
47+
48+
**After:**
49+
```
50+
$ pip install google-cloud-bigquery
51+
```
52+
53+
54+
## Type Annotations
55+
56+
The library is now type-annotated and declares itself as such. If you use a static
57+
type checker such as `mypy`, you might start getting errors in places where
58+
`google-cloud-bigquery` package is used.
59+
60+
It is recommended to update your code and/or type annotations to fix these errors, but
61+
if this is not feasible in the short term, you can temporarily ignore type annotations
62+
in `google-cloud-bigquery`, for example by using a special `# type: ignore` comment:
63+
64+
```py
65+
from google.cloud import bigquery # type: ignore
66+
```
67+
68+
But again, this is only recommended as a possible short-term workaround if immediately
69+
fixing the type check errors in your project is not feasible.
70+
71+
## Re-organized Types
72+
73+
The auto-generated parts of the library has been removed, and proto-based types formerly
74+
found in `google.cloud.bigquery_v2` have been replaced by the new implementation (but
75+
see the [section](#legacy-types) below).
76+
77+
For example, the standard SQL data types should new be imported from a new location:
78+
79+
**Before:**
80+
```py
81+
from google.cloud.bigquery_v2 import StandardSqlDataType
82+
from google.cloud.bigquery_v2.types import StandardSqlField
83+
from google.cloud.bigquery_v2.types.standard_sql import StandardSqlStructType
84+
```
85+
86+
**After:**
87+
```py
88+
from google.cloud.bigquery import StandardSqlDataType
89+
from google.cloud.bigquery.standard_sql import StandardSqlField
90+
from google.cloud.bigquery.standard_sql import StandardSqlStructType
91+
```
92+
93+
The `TypeKind` enum defining all possible SQL types for schema fields has been renamed
94+
and is not nested anymore under `StandardSqlDataType`:
95+
96+
97+
**Before:**
98+
```py
99+
from google.cloud.bigquery_v2 import StandardSqlDataType
100+
101+
if field_type == StandardSqlDataType.TypeKind.STRING:
102+
...
103+
```
104+
105+
**After:**
106+
```py
107+
108+
from google.cloud.bigquery import StandardSqlTypeNames
109+
110+
if field_type == StandardSqlTypeNames.STRING:
111+
...
112+
```
113+
114+
115+
## Issuing queries with `Client.create_job` preserves destination table
116+
117+
The `Client.create_job` method no longer removes the destination table from a
118+
query job's configuration. Destination table for the query can thus be
119+
explicitly defined by the user.
120+
121+
122+
## Changes to data types when reading a pandas DataFrame
123+
124+
The default dtypes returned by the `to_dataframe` method have changed.
125+
126+
* Now, the BigQuery `BOOLEAN` data type maps to the pandas `boolean` dtype.
127+
Previously, this mapped to the pandas `bool` dtype when the column did not
128+
contain `NULL` values and the pandas `object` dtype when `NULL` values are
129+
present.
130+
* Now, the BigQuery `INT64` data type maps to the pandas `Int64` dtype.
131+
Previously, this mapped to the pandas `int64` dtype when the column did not
132+
contain `NULL` values and the pandas `float64` dtype when `NULL` values are
133+
present.
134+
* Now, the BigQuery `DATE` data type maps to the pandas `dbdate` dtype, which
135+
is provided by the
136+
[db-dtypes](https://googleapis.dev/python/db-dtypes/latest/index.html)
137+
package. If any date value is outside of the range of
138+
[pandas.Timestamp.min](https://pandas.pydata.org/docs/reference/api/pandas.Timestamp.min.html)
139+
(1677-09-22) and
140+
[pandas.Timestamp.max](https://pandas.pydata.org/docs/reference/api/pandas.Timestamp.max.html)
141+
(2262-04-11), the data type maps to the pandas `object` dtype. The
142+
`date_as_object` parameter has been removed.
143+
* Now, the BigQuery `TIME` data type maps to the pandas `dbtime` dtype, which
144+
is provided by the
145+
[db-dtypes](https://googleapis.dev/python/db-dtypes/latest/index.html)
146+
package.
147+
148+
149+
## Changes to data types loading a pandas DataFrame
150+
151+
In the absence of schema information, pandas columns with naive
152+
`datetime64[ns]` values, i.e. without timezone information, are recognized and
153+
loaded using the `DATETIME` type. On the other hand, for columns with
154+
timezone-aware `datetime64[ns, UTC]` values, the `TIMESTAMP` type is continued
155+
to be used.
156+
157+
## Changes to `Model`, `Client.get_model`, `Client.update_model`, and `Client.list_models`
158+
159+
The types of several `Model` properties have been changed.
160+
161+
- `Model.feature_columns` now returns a sequence of `google.cloud.bigquery.standard_sql.StandardSqlField`.
162+
- `Model.label_columns` now returns a sequence of `google.cloud.bigquery.standard_sql.StandardSqlField`.
163+
- `Model.model_type` now returns a string.
164+
- `Model.training_runs` now returns a sequence of dictionaries, as recieved from the [BigQuery REST API](https://cloud.google.com/bigquery/docs/reference/rest/v2/models#Model.FIELDS.training_runs).
165+
166+
<a name="legacy-protobuf-types"></a>
167+
## Legacy Protocol Buffers Types
168+
169+
For compatibility reasons, the legacy proto-based types still exists as static code
170+
and can be imported:
171+
172+
```py
173+
from google.cloud.bigquery_v2 import Model # a sublcass of proto.Message
174+
```
175+
176+
Mind, however, that importing them will issue a warning, because aside from
177+
being importable, these types **are not maintained anymore**. They may differ
178+
both from the types in `google.cloud.bigquery`, and from the types supported on
179+
the backend.
180+
181+
### Maintaining compatibility with `google-cloud-bigquery` version 2.0
182+
183+
If you maintain a library or system that needs to support both
184+
`google-cloud-bigquery` version 2.x and 3.x, it is recommended that you detect
185+
when version 2.x is in use and convert properties that use the legacy protocol
186+
buffer types, such as `Model.training_runs`, into the types used in 3.x.
187+
188+
Call the [`to_dict`
189+
method](https://proto-plus-python.readthedocs.io/en/latest/reference/message.html#proto.message.Message.to_dict)
190+
on the protocol buffers objects to get a JSON-compatible dictionary.
191+
192+
```py
193+
from google.cloud.bigquery_v2 import Model
194+
195+
training_run: Model.TrainingRun = ...
196+
training_run_dict = training_run.to_dict()
197+
```
18198

19199
# 2.0.0 Migration Guide
20200

‎docs/index.rst‎

Lines changed: 2 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -30,7 +30,8 @@ API Reference
3030
Migration Guide
3131
---------------
3232

33-
See the guide below for instructions on migrating to the 2.x release of this library.
33+
See the guides below for instructions on migrating from older to newer *major* releases
34+
of this library (from ``1.x`` to ``2.x``, or from ``2.x`` to ``3.x``).
3435

3536
.. toctree::
3637
:maxdepth: 2

0 commit comments

Comments
 (0)