sql.h not found when installing PyODBC on Heroku

Question

I'm trying to install PyODBC on Heroku, but I get fatal error: sql.h: No such file or directory in the logs when pip runs. How do I fix this error?

Answer 1

To follow up on the answer below...

Example for Ubuntu:

sudo apt-get install unixodbc unixodbc-dev

Example for CentOS:

sudo yum install unixODBC-devel

On Windows:

conn = pyodbc.connect('DRIVER={SQL Server};SERVER=yourserver.yourcompany.com;DATABASE=yourdb;UID=user;PWD=password')

On Linux:

conn = pyodbc.connect('DRIVER={FreeTDS};SERVER=yourserver.yourcompany.com;PORT=1433;DATABASE=yourdb;UID=user;PWD=password;TDS_VERSION=7.2')

Answer 2

You need the unixODBC devel package. I don't know what distro you are using but you can google it and build from source.

Answer 3

You can add Heroku build pack to preinstall required apt packages first

heroku buildpacks:add --index 1 https://github.com/heroku/heroku-buildpack-apt

Add Aptfile in the your directory root and to the repository as well

unixodbc
unixodbc-dev
python-pyodbc
libsqliteodbc

It will install everything you need to work with pyodbc or aioodbc packages from python on Heroku

Answer 4

Create a file named Aptfile in the root directory of your project. The content of the file should be:

unixodbc
unixodbc-dev
python-pyodbc

Pass this command on the Command Prompt/shell:

**heroku buildpacks:add --index 1 https://github.com/heroku/heroku-buildpack-apt**

After completing above steps, commit the repository and then push it using

git push heroku master

Answer 5

The other answers are more or less correct; you're missing the unixodbc-dev[el] package for your operating system; that's what pip needs in order to build pyodbc from source.

However, a much easier option is to install pyodbc via the system package manager. On Debian/Ubuntu, for example, that would be apt-get install python-pyodbc. Since pyodbc has a lot of compiled components and interfaces heavily with the UnixODBC OS-level packages, it is probably a better fit for a system package rather than a Python/pip-installed one.

You can still list it as a dependency in your requirements.txt files if you're making code for distribution, but it'll usually be easier to install it via the system PM.