NarayanB edited a comment on issue #12416:
URL: https://github.com/apache/arrow/issues/12416#issuecomment-1042969321
```
import pandas as pd
import polars as pl
import pyarrow as pa
import pyarrow.parquet as pq
import pyarrow.dataset as ds
def read_csv(csv_file):
print("reading csv")
csv = """pid1,pid2,intCol,strCol
2010,10000,23455555508999,Peaceful
2015,15000,7753285016841556620,Happy
2020,25000,,World""".encode()
good_df = (pl.read_csv(csv, dtypes={"intCol": pl.Utf8})
.with_column(pl.col("intCol").str.replace("", "0").cast(pl.Int64))
)
bad_df = pl.read_csv(csv)
df = good_df
#df = bad_df
print(df.head(10))
table = df.to_arrow()
print('Table Schema..\n',table.schema)
return table
def save_table(table, location):
pq.write_to_dataset(table, location, partition_cols=['pid1','pid2'])
def read_table(location):
schema = pa.schema([ ('pid1', pa.int64()),
('pid2', pa.int64())
])
partition = ds.partitioning(schema=schema, flavor='hive')
dataset = ds.dataset(location, partitioning=partition)
table = dataset.to_table()
print("Retrived table schema\n", table)
df = pl.from_arrow(table)
print(df.head(10))
table = read_csv(None)
save_table(table, '../data')
read_table('../data')
```
Please run this program twice once with df = good_df and next time df =
bad_df. You will see the issue that the int64 value toggling between
7753285016841556620 vs 7753285016841556992. You will have to clear the
partition data for each run
I think I issue is that for some very big int64 values having null in one
field column creates the issue. There is no pandas involved here. I haven't
seen the issue if the int64 field value is something like 10000 or 2500001 etc
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
