[jira] [Commented] (ARROW-4032) [Python] New pyarrow.Table.from_pylist() function

["David Lee (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/ARROW-4032?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16721986#comment-16721986 ] 

David Lee commented on ARROW-4032:
----------------------------------

Ended up just writing from_pylist() and to_pylist().. They run much faster than going through pandas..
{code:java}
def from_pylist(pylist, schema, safe=True):
    arrow_columns = list()
    for column in schema.names:
        arrow_columns.append(pa.array([v[column] if column in v else None for v in pylist], safe=safe, type=schema.types[schema.get_field_index(column)]))
    arrow_table = pa.Table.from_arrays(arrow_columns, columns)
    return arrow_table

def to_pylist(arrow_table):
    od = pyarrow.Table.to_pydict(arrow_table)
    pylist = list()
    columns = list(arrow_table.keys())
    rows = len(arrow_table[columns[0]])
    for row in range(rows):
        pylist.append({key: arrow_table[key][row] for key in columns})
    return pylist
{code}

> [Python] New pyarrow.Table.from_pylist() function
> -------------------------------------------------
>
>                 Key: ARROW-4032
>                 URL: https://issues.apache.org/jira/browse/ARROW-4032
>             Project: Apache Arrow
>          Issue Type: Task
>          Components: Python
>            Reporter: David Lee
>            Priority: Minor
>
> Here's a proposal to create a pyarrow.Table.from_pydict() function.
> Right now only pyarrow.Table.from_pandas() exist and there are inherit problems using Pandas with NULL support for Int(s) and Boolean(s)
> [http://pandas.pydata.org/pandas-docs/version/0.23.4/gotchas.html]
> {{NaN}}, Integer {{NA}} values and {{NA}} type promotions:
> Sample python code on how this would work.
>  
> {code:java}
> import pyarrow as pa
> from datetime import datetime
> # convert microseconds to milliseconds. More support for MS in parquet.
> today = datetime.now()
> today = datetime(today.year, today.month, today.day, today.hour, today.minute, today.second, today.microsecond - today.microsecond % 1000)
> test_list = [
> {"name": "Tom", "age": 10},
> {"name": "Mark", "age": 5, "city": "San Francisco"},
> {"name": "Pam", "age": 7, "birthday": today}
> ]
> def from_pylist(pylist, schema=None, columns=None, safe=True):
>     arrow_columns = list()
>     if schema:
>         columns = schema.names
>     if not columns:
>         return
>     for column in columns:
>         arrow_columns.append(pa.array([v[column] if column in v else None for v in pylist], safe=safe))
>     arrow_table = pa.Table.from_arrays(arrow_columns, columns)
>     if schema:
>         arrow_table = arrow_table.cast(schema, safe=safe)
>     return arrow_table
> test = from_pylist(test_list, columns=['name' , 'age', 'city', 'birthday', 'dummy'])
> test_schema = pa.schema([
> pa.field('name', pa.string()),
> pa.field('age', pa.int16()),
> pa.field('city', pa.string()),
> pa.field('birthday', pa.timestamp('ms'))
> ])
> test2 = from_pylist(test_list, schema=test_schema)
> {code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)