Build web application with Flask on Heroku.

Table of content

Heroku deployment

Create an heroku instance with a predefined name.
```
$heroku create sentimentx
```
Create a virtual environment with the virtualenv venv which will generate the information as follows.
```
New python executable in venv/bin/python
Installing setuptools, pip, wheel...done.
```

Install python packages with pip for the virtual environment we just created.

$pip install tweepy
$pip install flask
$pip install numpy
$pip install cPickle
$pip install nltk
$pip install scipy

Still, we need to have a Python dependency file requirements.txt for Heroku to understand these required packages. As all necessary packages are installed with pip, we can collect package information with freeze shown as follows.
```
$pip freeze > requirements.txt
```
It seems that packages are in good shape. However, there are some problems with Heroku to install scipy and numpy and probably scikit-learn. In practice, it means that a application with these packages can run nicely on local computer but won’t run on Heroku remote server. The solution is to use a third party pre-build package with the following command. It is worth noting that the following pre-build package at least support numpy==1.8.1 and scipy==0.14.0. So if you have a higher version of these two packages, you might want to consider lower version alternatives in case they don’t work with Heroku.
```
heroku buildpacks:set https://github.com/thenovices/heroku-buildpack-scipy
```
Write the Procfile with the following content to tell Heroku server which python script should be activated.
```
web: python app.py
```
Deploy the web application to heroku with the following command. The command will first copy all files to heroku server and deploy them afterwords. Of course, you need to have them ready already in Github e.g., all related files have been committed to Github local/remote.
```
$git push heroku master
```
To check if the submitted web services is running on Heroku remote service, you can use the following command.
```
$heroku ps:scale web=1
```
Or open a browser window to the web service running in Heroku remote server with the following command.
```
$heroku open
```
Or run Heroku web services on local machine with the following command.
```
$heroku local
```
In case of problems that the web application did not run on heroku server check log with the following command.
```
$heroku logs
```
It is also nice to know some limit of the Heroku server.

Build web applications with Flask+Heroku

Table of content

Heroku deployment