installable on Raspberry Pi? about fastparquet HOT 9 CLOSED

Numba does not yet support the RPi3 (armv7 using the typical raspbian install), although we have a PR to add support here:

numba/numba#1968

As noted in the PR discussion, not all of our unit tests pass on ARMv7 for some reason.

from fastparquet.

I also have newer pandas for ARM in my Anaconda Cloud channel:

https://anaconda.org/seibert/pandas

from fastparquet.

I believe that all necessary packages run on ARM. @seibert would know more.

from fastparquet.

snappy is not a hard dependency, but without it, you can't use that compression, of course.

from fastparquet.

If you get fastparquet working on a raspberryPi, I for one definitely want to know your use case!

from fastparquet.

Installed Miniconda on my pi3...

then conda install -c conda-forge fastparquet...no armv6 packet okay
with pip install git+https://github.com/dask/fastparquet i get a issue about the llvmlite modul...

conda install llvmlite solve this problem.

....more to come

from fastparquet.

@rddaz2013 , any success? Would be interested to know your use case.

from fastparquet.

1. Fresh PI-3
2. Install Mini-conda + requirements

pi@raspberrypi:~/fastparquet/fastparquet/test $ python            
Python 3.4.3 |Continuum Analytics, Inc.| (default, Aug 21 2015, 00:53:08) 
[GCC 4.6.3] on linux

3. Modify fastparquet to work without Numba...

4. Problem the mini-conda pandas for arm is Version 0.16.2

> pi@raspberrypi:~/fastparquet/fastparquet/test $ python            
Python 3.4.3 |Continuum Analytics, Inc.| (default, Aug 21 2015, 00:53:08) 
[GCC 4.6.3] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pandas as pd
>>> pd.__version__
'0.16.2'

with this i get an error running test_api.py

pi@raspberrypi:~/fastparquet/fastparquet/test $ python test_api.py
Traceback (most recent call last):
  File "test_api.py", line 8, in <module>
    from fastparquet.util import tempdir
  File "/home/pi/miniconda3/lib/python3.4/site-packages/fastparquet-0.0.3-py3.4.egg/fastparquet/__init__.py", line 16, in <module>
    from .writer import write
  File "/home/pi/miniconda3/lib/python3.4/site-packages/fastparquet-0.0.3-py3.4.egg/fastparquet/writer.py", line 21, in <module>
    from . import encoding, api
  File "/home/pi/miniconda3/lib/python3.4/site-packages/fastparquet-0.0.3-py3.4.egg/fastparquet/api.py", line 18, in <module>
    from . import core, schema, converted_types, encoding, writer, dataframe
  File "/home/pi/miniconda3/lib/python3.4/site-packages/fastparquet-0.0.3-py3.4.egg/fastparquet/dataframe.py", line 6, in <module>
    from pandas.core.index import RangeIndex, Index
ImportError: cannot import name 'RangeIndex'

RangeIndex -> pandas v 0.18+ ??

from fastparquet.

" I for one definitely want to know your use case! "

1. Proof of Concept for my 3 Working system..(Win, Linux and the PI)
2. Python + Scipy + Numpy + matplotlib ( today only python2.10 but work for python3 in progress)
3. temperature/pressure records with time intervall from 1/10 ms to minutes + duration of measurement of minutes or month.
4. raspberry pi = independent data collector for slow data and event logging.
5. At the moment the source data is..first stage TXT! (~ up to 200 MB) second stage Numpy.arrays... works for me, but is not a standardized data format.

Not really BIG-Data, but too much for Excel .-)

from fastparquet.