Read Avro file from Pandas

avro_to_dataframe.py

import pandas
import fastavro

def avro_df(filepath, encoding):
    # Open file stream
    with open(filepath, encoding) as fp:
        # Configure Avro reader
        reader = fastavro.reader(fp)
        # Load records in memory
        records = [r for r in reader]
        # Populate pandas.DataFrame with records
        df = pandas.DataFrame.from_records(records)
        # Return created DataFrame
        return df

👍

What will be the encoding?

What will be the encoding?

"Encoding" (actually, referring to the python mode of opening the file) should be 'rb' .
This is a typo, should be named "mode" instead of "encoding".

Thanks, very useful

Hi thanks !

But when we are printing df,data is displayed in dictionary format. is there any way we can covert in tabular format ?

As of i am getting data like below example :

IPython 7.8.0 -- An enhanced Interactive Python.
<bound method DataFrame.count of MbrData ... Identifiers
0 {'SourceName': 'AFFINITY', 'SourceId': '159234... ... [{'Identifiers': 'DUNSNumber', 'Identifier_Val...
1 {'SourceName': 'AFFINITY', 'SourceId': '595713... ... [{'Identifiers': 'DUNSNumber', 'Identifier_Val...
2 {'SourceName': 'AFFINITY', 'SourceId': '168155... ... [{'Identifiers': 'DUNSNumber', 'Identifier_Val...
3 {'SourceName': 'AFFINITY', 'SourceId': '725398... ... [{'Identifiers': 'DUNSNumber', 'Identifier_Val...
4 {'SourceName': 'AFFINITY', 'SourceId': '727384... ... [{'Identifiers': 'DUNSNumber', 'Identifier_Val...

Hey @Neetu2407 ,
I don't know much about avro files, but you can check if the avro file schema is parsed correctly with print(reader.schema)

You can look for more in https://www.perfectlyrandom.org/2019/11/29/handling-avro-files-in-python/

Also, instead of printing df, df.head() and df.dtypes may help.
Is the dataset public? Can we at least see your schema and corresponding df.columns? If you want me to take a look, drop me an email.

I had an error 'utf8' codec can't decode byte 0x83. This was solved by using open(filepath, 'rb'), where the b means to read the file in binary format. As already mentioned, this argument is the "mode", not "encoding".

It looks like you can save a line of code and avoid temporarily duplicating the data in memory by passing the reader iterable directly to from_records rather than loading it into a list first.