git-history 0.3.1

Tools for analyzing Git history using SQLite

git-history

Tools for analyzing Git history using SQLite

Installation

Install this tool using pip:

$ pip install git-history

Usage

This tool can be run against a Git repository that holds a file that contains JSON, CSV/TSV or some other format and which has multiple versions tracked in the Git history. See Git scraping to understand how you might create such a repository.

The file command analyzes the history of an individual file within the repository, and generates a SQLite database table that represents the different versions of that file over time.

The file is assumed to contain multiple objects - for example, the results of scraping an electricity outage map or a CSV file full of records.

Assuming you have a file called incidents.json that is a JSON array of objects, with multiple versions of that file recorded in a repository.

Change directory into the GitHub repository in question and run the following:

git-convert file incidents.db incidents.json

This will create a new SQLite database in the incidents.db file with two tables:

• commits containing a row for every commit, with a hash column and the commit_at date.
• items containing a row for every item in every version of the filename.json file - with an extra commit column that is a foreign key back to the commits table.

If you have 10 historic versions of the incidents.json file and each one contains 30 incidents, you will end up with 10 * 30 = 300 rows in your items table.

De-duplicating items using IDs

If your objects have a unique identifier - or multiple columns that together form a unique identifier - you can use the --id option to de-duplicate and track changes to each of those items over time.

If there is a unique identifier column called IncidentID you could run the following:

git-convert file incidents.db incidents.json --id IncidentID

This will create three tables - commits, items and item_versions.

The items table will contain just the most recent version of each row, de-duplicated by ID.

The item_versions table will contain a row for each captured differing version of that item, plus the following columns:

• item as a foreign key to the items table
• commit as a foreign key to the commits table
• version as the numeric version number, starting at 1 and incrementing for each captured version

If you have already imported history, the command will skip any commits that it has seen already and just process new ones. This means that even though an initial import could be slow subsequent imports should run a lot faster.

Additional options:

• --repo DIRECTORY - the path to the Git repository, if it is not the current working directory.
• --branch TEXT - the Git branch to analyze - defaults to main.
• --id TEXT - as described above: pass one or more columns that uniquely identify a record, so that changes to that record can be calculated over time.
• --ignore TEXT - one or more columns to ignore - they will not be included in the resulting database.
• --csv - treat the data is CSV or TSV rather than JSON, and attempt to guess the correct dialect
• --convert TEXT - custom Python code for a conversion, see below.
• --import TEXT - Python modules to import for --convert.
• --ignore-duplicate-ids - if a single version of a file has the same ID in it more than once, the tool will exit with an error. Use this option to ignore this and instead pick just the first of the two duplicates.
• --silent - don't show the progress bar.

Note that id, item, version, commit and rowid are reserved column names that are used by this tool. If your data contains any of these they will be renamed to id_, item_, version_, commit_ or rowid_ to avoid clashing with the reserved columns.

There is one exception: if you have an id column and use --id id without specifying more than one ID column, your ìd` column will be used as the item ID but will not be renamed.

CSV and TSV data

If the data in your repository is a CSV or TSV file you can process it by adding the --csv option. This will attempt to detect which delimiter is used by the file, so the same option works for both comma- and tab-separated values.

git-convert file trees.db trees.csv --id TreeID

Custom conversions using --convert

If your data is not already either CSV/TSV or a flat JSON array, you can reshape it using the --convert option.

The format needed by this tool is an array of dictionaries that looks like this:

[
    {
        "id": "552",
        "name": "Hawthorne Fire",
        "engines": 3
    },
    {
        "id": "556",
        "name": "Merlin Fire",
        "engines": 1
    }
]

If your data does not fit this shape, you can provide a snippet of Python code to converts the on-disk content of each stored file into a Python list of dictionaries.

For example, if your stored files each look like this:

{
    "incidents": [
        {
            "id": "552",
            "name": "Hawthorne Fire",
            "engines": 3
        },
        {
            "id": "556",
            "name": "Merlin Fire",
            "engines": 1
        }
    ]
}

You could use the following Python snippet to convert them to the required format:

json.loads(content)["incidents"]

(The json module is exposed to your custom function by default.)

You would then run the tool like this:

git-convert file database.db incidents.json \
  --id id \
  --convert 'json.loads(content)["incidents"]'

The content variable is always a bytes object representing the content of the file at a specific moment in the repository's history.

You can import additional modules using --import. This example shows how you could read a CSV file that uses ; as the delimiter:

git-history file trees.db ../sf-tree-history/Street_Tree_List.csv \
  --repo ../sf-tree-history \
  --import csv \
  --import io \
  --convert '
    fp = io.StringIO(content.decode("utf-8"))
    return list(csv.DictReader(fp, delimiter=";"))
    ' \
  --id TreeID

If your Python code spans more than one line it needs to include a return statement.

Development

To contribute to this tool, first checkout the code. Then create a new virtual environment:

cd git-history
python -m venv venv
source venv/bin/activate

Or if you are using pipenv:

pipenv shell

Now install the dependencies and test dependencies:

pip install -e '.[test]'

To run the tests:

pytest