How to read a list of parquet files from S3 as a pandas dataframe using pyarrow?

I have a hacky way of achieving this using boto3 (1.4.4), pyarrow (0.4.1) and pandas (0.20.3).

First, I can read a single parquet file locally like this:

import pyarrow.parquet as pq

path = 'parquet/part-r-00000-1e638be4-e31f-498a-a359-47d017a0059c.gz.parquet'
table = pq.read_table(path)
df = table.to_pandas()

I can also read a directory of parquet files locally like this:

import pyarrow.parquet as pq

dataset = pq.ParquetDataset('parquet/')
table = dataset.read()
df = table.to_pandas()

Both work like a charm. Now I want to achieve the same remotely with files stored in a S3 bucket. I was hoping that something like this would work:

dataset = pq.ParquetDataset('s3n://dsn/to/my/bucket')

But it does not:

OSError: Passed non-file path: s3n://dsn/to/my/bucket

After reading pyarrow's documentation thoroughly, this does not seem possible at the moment. So I came out with the following solution:

Reading a single file from S3 and getting a pandas dataframe:

import io
import boto3
import pyarrow.parquet as pq

buffer = io.BytesIO()
s3 = boto3.resource('s3')
s3_object = s3.Object('bucket-name', 'key/to/parquet/file.gz.parquet')
s3_object.download_fileobj(buffer)
table = pq.read_table(buffer)
df = table.to_pandas()

And here my hacky, not-so-optimized, solution to create a pandas dataframe from a S3 folder path:

import io
import boto3
import pandas as pd
import pyarrow.parquet as pq

bucket_name = 'bucket-name'
def download_s3_parquet_file(s3, bucket, key):
    buffer = io.BytesIO()
    s3.Object(bucket, key).download_fileobj(buffer)
    return buffer

client = boto3.client('s3')
s3 = boto3.resource('s3')
objects_dict = client.list_objects_v2(Bucket=bucket_name, Prefix='my/folder/prefix')
s3_keys = [item['Key'] for item in objects_dict['Contents'] if item['Key'].endswith('.parquet')]
buffers = [download_s3_parquet_file(s3, bucket_name, key) for key in s3_keys]
dfs = [pq.read_table(buffer).to_pandas() for buffer in buffers]
df = pd.concat(dfs, ignore_index=True)

Is there a better way to achieve this? Maybe some kind of connector for pandas using pyarrow? I would like to avoid using pyspark, but if there is no other solution, then I would take it.

You should use the s3fs module as proposed by yjk21. However as result of calling ParquetDataset you'll get a pyarrow.parquet.ParquetDataset object. To get the Pandas DataFrame you'll rather want to apply .read_pandas().to_pandas() to it:

import pyarrow.parquet as pq
import s3fs
s3 = s3fs.S3FileSystem()

pandas_dataframe = pq.ParquetDataset('s3://your-bucket/', filesystem=s3).read_pandas().to_pandas()

Thanks! Your question actually tell me a lot. This is how I do it now with pandas (0.21.1), which will call pyarrow, and boto3 (1.3.1).

import boto3
import io
import pandas as pd

# Read single parquet file from S3
def pd_read_s3_parquet(key, bucket, s3_client=None, **args):
    if s3_client is None:
        s3_client = boto3.client('s3')
    obj = s3_client.get_object(Bucket=bucket, Key=key)
    return pd.read_parquet(io.BytesIO(obj['Body'].read()), **args)

# Read multiple parquets from a folder on S3 generated by spark
def pd_read_s3_multiple_parquets(filepath, bucket, s3=None, 
                                 s3_client=None, verbose=False, **args):
    if not filepath.endswith('/'):
        filepath = filepath + '/'  # Add '/' to the end
    if s3_client is None:
        s3_client = boto3.client('s3')
    if s3 is None:
        s3 = boto3.resource('s3')
    s3_keys = [item.key for item in s3.Bucket(bucket).objects.filter(Prefix=filepath)
               if item.key.endswith('.parquet')]
    if not s3_keys:
        print('No parquet found in', bucket, filepath)
    elif verbose:
        print('Load parquets:')
        for p in s3_keys: 
            print(p)
    dfs = [pd_read_s3_parquet(key, bucket=bucket, s3_client=s3_client, **args) 
           for key in s3_keys]
    return pd.concat(dfs, ignore_index=True)

Then you can read multiple parquets under a folder from S3 by

df = pd_read_s3_multiple_parquets('path/to/folder', 'my_bucket')

(One can simplify this code a lot I guess.)

It can be done using boto3 as well without the use of pyarrow

import boto3
import io
import pandas as pd

# Read the parquet file
buffer = io.BytesIO()
s3 = boto3.resource('s3')
object = s3.Object('bucket_name','key')
object.download_fileobj(buffer)
df = pd.read_parquet(buffer)

print(df.head())

Probably the easiest way to read parquet data on the cloud into dataframes is to use dask.dataframe in this way:

import dask.dataframe as dd
df = dd.read_parquet('s3://bucket/path/to/data-*.parq')

dask.dataframe can read from Google Cloud Storage, Amazon S3, Hadoop file system and more!

Provided you have the right package setup

$ pip install pandas==1.1.0 pyarrow==1.0.0 s3fs==0.4.2

and your AWS shared config and credentials files configured appropriately

you can use pandas right away:

import pandas as pd

df = pd.read_parquet("s3://bucket/key.parquet")

---

In case of having multiple AWS profiles you may also need to set

$ export AWS_DEFAULT_PROFILE=profile_under_which_the_bucket_is_accessible

so you can access your bucket.

If you are open to also use AWS Data Wrangler.

import awswrangler as wr

df = wr.s3.read_parquet(path="s3://...")

You can use s3fs from dask which implements a filesystem interface for s3. Then you can use the filesystem argument of ParquetDataset like so:

import s3fs
s3 = s3fs.S3FileSystem()
dataset = pq.ParquetDataset('s3n://dsn/to/my/bucket', filesystem=s3)