Undefined symbol in turbodbc_arrow_support.so - pip install on Debian

[peter-wimsey]

Hi,

I am trying to build a docker image of turbodbc with Arrow support but get the following error:

Python 3.7.3 (default, Jun 11 2019, 01:05:09) 
[GCC 6.3.0 20170516] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import turbodbc_arrow_support
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: /usr/local/lib/python3.7/site-packages/turbodbc_arrow_support.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZN5arrow6StatusC1ENS_10StatusCodeERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE

Any ideas what's going wrong here? Some more info below.

Thanks

---

# pip list
Package    Version
---------- -------
numpy      1.16.4 
pip        19.1.1 
pyarrow    0.13.0 
pybind11   2.3.0  
setuptools 41.0.1 
six        1.12.0 
turbodbc   3.1.1  
wheel      0.33.4 

---

Dockerfile:

FROM python:3.7

USER root

ARG http_proxy

RUN export http_proxy=$http_proxy https_proxy=$http_proxy \
    && apt-get update \
    && apt-get install -y build-essential libboost-all-dev unixodbc-dev python3-dev

RUN export http_proxy=$http_proxy https_proxy=$http_proxy \
    && pip install numpy pyarrow turbodbc

[davlee1972]

I'm seeing the same error with Turbodbc 3.2.0 and Python 3.6. I was able to successfully build turbodbc using conda, but it was giving me a No Arrow Extension support error when trying to do a "arrow_table = cursor.fetchallarrow()" call. I traced the error to "import turbodbc_arrow_support" failure as well.

ImportError: /usr/local/Anaconda3-5.3.1-Linux-x86_64/envs/pyspark-3-6/lib/python3.6/site-packages/turbodbc_arrow_support.cpython-36m-x86_64-linux-gnu.so: undefined symbol: _ZN5arrow6StatusC1ENS_10StatusCodeERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE

[davlee1972]

Python 3.6.9 |Anaconda, Inc.| (default, Jul 30 2019, 19:07:31)
[GCC 7.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pyarrow as pa
>>> import turbodbc_arrow_support
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ImportError: /usr/local/Anaconda3-5.3.1-Linux-x86_64/envs/pyspark-3-6/lib/python3.6/site-packages/turbodbc_arrow_support.cpython-36m-x86_64-linux-gnu.so: undefined symbol: _ZN5arrow6StatusC1ENS_10StatusCodeERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE

[davlee1972]

Got it working. I think the problem was that I pulled pybind11 and six from pip instead of conda as part of the turbodbc install when running "pip install turbodbc" in a conda environment

[xhochy]

@davlee1972 When running in a conda environment, you should always install turbodbc with conda install turbodbc -c conda-forge. This should work much better than with using pip.

[peter-wimsey]

Thanks @davlee1972, installing with conda did indeed fix the issue.

[alexkim-avant]

What do you recommend if I can only use pip to install turbodbc?