How to read a Parquet file into Pandas DataFrame?

Question

How to read a modestly sized Parquet data-set into an in-memory Pandas DataFrame without setting up a cluster computing infrastructure such as Hadoop or Spark? This is only a moderate amount of data that I would like to read in-memory with a simple Python script on a laptop. The data does not reside on HDFS. It is either on the local file system or possibly in S3. I do not want to spin up and configure other services like Hadoop, Hive or Spark.

I thought Blaze/Odo would have made this possible: the Odo documentation mentions Parquet, but the examples seem all to be going through an external Hive runtime.

Solution 1

pandas 0.21 introduces new functions for Parquet:

import pandas as pd
pd.read_parquet('example_pa.parquet', engine='pyarrow')

or

import pandas as pd
pd.read_parquet('example_fp.parquet', engine='fastparquet')

The above link explains:

These engines are very similar and should read/write nearly identical parquet format files. These libraries differ by having different underlying dependencies (fastparquet by using numba, while pyarrow uses a c-library).

Solution 2

Update: since the time I answered this there has been a lot of work on this look at Apache Arrow for a better read and write of parquet. Also: http://wesmckinney.com/blog/python-parquet-multithreading/

There is a python parquet reader that works relatively well: https://github.com/jcrobak/parquet-python

It will create python objects and then you will have to move them to a Pandas DataFrame so the process will be slower than pd.read_csv for example.

Solution 3

Aside from pandas, Apache pyarrow also provides way to transform parquet to dataframe

The code is simple, just type:

import pyarrow.parquet as pq

df = pq.read_table(source=your_file_path).to_pandas()

For more information, see the document from Apache pyarrow Reading and Writing Single Files

Solution 4

Parquet

Step 1: Data to play with

df = pd.DataFrame({
    'student': ['personA007', 'personB', 'x', 'personD', 'personE'],
    'marks': [20,10,22,21,22],
})

Step 2: Save as Parquet

df.to_parquet('sample.parquet')

Step 3: Read from Parquet

df = pd.read_parquet('sample.parquet')

Solution 5

Considering the .parquet file named data

parquet_file = '../data.parquet'

open( parquet_file, 'w+' )

Then use pandas.to_parquet (this function requires either the fastparquet or pyarrow library)

parquet_df.to_parquet(parquet_file)

Then, use pandas.read_parquet() to get a dataframe

new_parquet_df = pd.read_parquet(parquet_file)

View more solutions

Comments

[mdurant]

Do you happen to have the data openly available? My branch of python-parquet github.com/martindurant/parquet-python/tree/py3 had a pandas reader in parquet.rparquet, you could try it. There are many parquet constructs it cannot handle.

[XValidated]

Wait for the Apache Arrow project that the Pandas author Wes Mckinney is part of. wesmckinney.com/blog/pandas-and-apache-arrow After it is done, users should be able to read in Parquet file directly from Pandas.

[sroecker]

Since the question is closed as off-topic (but still the first result on Google) I have to answer in a comment.. You can now use pyarrow to read a parquet file and convert it to a pandas DataFrame: import pyarrow.parquet as pq; df = pq.read_table('dataset.parq').to_pandas()

[user48956]

Kinda annoyed that this question was closed. Spark and parquet are (still) relatively poorly documented. Am also looking for the answer to this.

[asmaier]

Have a look at github.com/dask/fastparquet . For an introduction see continuum.io/blog/developer-blog/introducing-fastparquet .

[ogrisel]

Both the fastparquet and pyarrow libraries make it possible to read a parquet file into a pandas dataframe: github.com/dask/fastparquet and arrow.apache.org/docs/python/parquet.html

[Andras Deak -- Слава Україні]

@ogrisel it's open now

[MichaelChirico]

@DanielMahler consider updating the accepted answer

[bluszcz]

Actually there is pyarrow which allows both reads / writes: pyarrow.readthedocs.io/en/latest/parquet.html

[snooze_bear]

I get a permission denied error when I try to follow your link, @bluszcz -- do you have an alternate?

[ogrisel]

parquet-python is much slower than alternatives such as fastparquet et pyarrow: arrow.apache.org/docs/python/parquet.html

[ogrisel]

pd.read_parquet is now part of pandas. The other answer should be marked as valid.

[Catbuilts]

For most of my data, 'fastparquet' is a bit faster. Just in case pd.read_parquet() returns a problem with Snappy Error, run conda install python-snappy to install snappy.

[Seb]

I found pyarrow to be too difficult to install (both on my local windows machine and on a cloud linux machine). Even after the python-snappy fix, there were additional issues with the compiler as well as the error module 'pyarrow' has no attribute 'compat'. fastparquet had no issues at all.

[Khan]

@Catbuilts You can use gzip if you don't have snappy.

[wawawa]

can 'fastparquet' read ',snappy.parquet' file?

[Mark Z.]

I had the opposite experience vs. @Seb. fastparquet had a bunch of issues, pyarrow was simple pip install and off I went