hi Igor -- you can use the Schema.serialize and pyarrow.read_schema methods
https://github.com/apache/arrow/blob/master/python/pyarrow/tests/test_ipc.py#L640 For short term storage you can also pickle schemas I didn't find read_schema in the API reference (http://arrow.apache.org/docs/python/api/ipc.html) so I opened a documentation JIRA https://issues.apache.org/jira/browse/ARROW-6201 - Wes On Fri, Aug 9, 2019 at 4:47 AM Игорь Ястребов <[email protected]> wrote: > > Hi everyone! > > Is there a recommended way to store schemata for Arrow tables on disk? I want > to load them later to provide information to csv reader (by constructing a > dictionary or directly if it gets implemented in the future). This is > necessary to read multiple csv files that follow the same origin but may get > wrong inferred type due to a lack of data in this particular file (null > fields, integer types instead of float types).
