dev tools

Here are listed some useful tools to know for development.

Git

Relative config files

Since Git 2.13, it is possible to define config files for directories.

Example: the file ~/dev/dbnomics define user email for all DBnomics repositories:

in ~/.gitconfig:

[includeIf "gitdir:~/dev/dbnomics/"]
    path = ~/dev/dbnomics/gitconfig

in ~/dev/dbnomics/gitconfig:
```
[user]
    email = jc.dus@cepremap.org
```

Multiple config files (one in the sub-tree of the other)

Example:

In ~/.gitconfig:

[includeIf "gitdir:~/dev/"]
    path = ~/dev/gitconfig
[includeIf "gitdir:~/dev/dbnomics/"]
    path = ~/dev/dbnomics/gitconfig

Forcing git to use a manually configured email address

In ~/.gitconfig:

[user]
    useConfigOnly = true

Delete local image of remote branches when fetching

After a merge request is approved and the branch merged, the branch is deleted on the server. On the machine of the developer, it is irrelevant to keep those polluting branches.

This setting makes git delete local branches when their remote branches were deleted on the server.

In ~/.gitconfig:

[fetch]
	prune = true

Good commit messages

https://chris.beams.io/posts/git-commit/

Python

Useful dev tools

jupyter-notebook

Virtualenvs

We recommend using virtualenv wrapper to ease shell integration
To automatically activate venvs: https://github.com/kennethreitz/autoenv

To use virtualenvs with Visual Studio Code editor:

from your terminal, while the virtualenv is activated, identify the path of the python executable:

which python
# Should answer something like /home/username/.local/share/virtualenvs/virtualenvname/bin/python

in Visual Studio Code, edit your workspace settings, and paste this block, adapting of course the path with your actual value:

{
    "python.pythonPath": "/home/username/.local/share/virtualenvs/virtualenvname/bin/python"
}

autopep8

We use autopep8 to reformat Python code automatically, and avoid bikeshedding.

It works very well with the Visual Studio Code editor, and is enabled by default, as soon as you pip install autopep8 in your venv.

First you must configure Visual Studio Code to work with virtualenvs (see above).

To define a limit of 120 characters per line, edit your user settings, and paste this block:

    "python.formatting.formatOnSave": true,
    "python.formatting.autopep8Args": [
        "--max-line-length=120"
    ],

pylint

We use pylint to check code quality directly from our source code editor.

It works very well with the Visual Studio Code editor.

First you must configure Visual Studio Code to work with virtualenvs (see above).

From your virtualenv in the shell, type pip install pylint.

By default you'll be annoyed by the huge amount of warnings, so follow the instructions on the Code Style page.

isort

To sort imports automatically and separate them in 3 sections (Python, third-party, local), use isort.

Install in your virtualenv:

pip install isort

To sort imports in Visual Studio Code, press F1 and execute "Sort imports" action.

CSV tools

Tabview

View big CSV files in terminal.

https://github.com/TabViewer/tabview

Xsv

Powerfull operations in CSV files made quickly.

https://github.com/BurntSushi/xsv

Sample

Usefull function: sample => to test a script parsing a HUGE CSV source file, instead of testing it on the n firsts lines of file, we can generate a sample of the file, representing a "rich" mix of the values found in the file:

    xsv sample 10000 services_annual_dataset.csv > services_annual_dataset-sample.csv

generate a 10000 lines sample of services_annual_dataset.csv file

Select & frequency

Ex: collect (and count) all the flags values of a file:

> xsv select Flag merchandise_indices_annual_dataset.csv | xsv frequency
field,value,count
Flag,(NULL),26860
Flag,E,9924
Flag,B,8

sshfs - mount remote directory from server

When git repositories are big, you may want to avoid cloning them on your development machine. Sometimes it is more convenient to access them remotely from the server.

Use sshfs as described here.

tmpfs - mount directory in RAM

When a directory grows in size of number of files, the disk I/O becomes slower, even with SSD disks.

A quick and dirty solution during development is to use tmpfs like so (paths are simplified):

cd json-data
sudo mount -t tmpfs -o size=8g tmpfs wto-json-data-tmpfs

cd wto-fetcher
./wto_to_dbnomics.py wto-source-data wto-json-data-tmpfs

cd json-data
sudo umount wto-json-data-tmpfs

Warning: after un-mounting, data is lost! Once your script is OK, execute it against the real wto-json-data directory.