This repository holds scripts that are useful for the creation of the NZSL Android and iOS applications.
An scheduled build runs monthly to produce the assets and data files for the NZSL Dictionary applications to use. If you are after these assets and are not interested in the absolute latest data, you can download these files without having to wait for them to build from the Releases tab.
Aside from scheduled releases, prerelease builds may occasionally be run. This build extracts sign data, but does not process assets like sign illustrations or videos. Prerelease data is not published as a Github release or as build artifacts, but is used to allow data to be checked in prerelease environments of our apps.
This tool contains a number of Github Actions which can be manually run by repository administrators to update the dictionary data used by NZSL Share and the NZSL Dictionary.
The deployment process is:
- Download the sign data from Signbank, and process it into a SQLite database
- Upload the SQLite database to an S3 bucket. This requires a role ARN, bucket name, and deployment ACL, which are specified as repository secrets and provisioned by https://github.com/odnzsl/nzsl-infrastructure.
- Restarts the Heroku applications (these applications download the latest sign data on boot using a presigned URL). This requires a valid Heroku app name and API token to be configured, which are specified as repository secrets.
- If the database being exported is a production release, the SQLite and .dat files produced are attached to the build as artifacts. Prerelease data is not available for download from the build and is limited for Prerelease app use (we use prereleases to perform a final check of data before publishing, but this sign data is not available for general use until this check is completed and the sign is published)
Configuration and secrets required for the deployment are managed by Ackama. If you are a Ackama staff member and require access, this may be requested by messaging the Codecare team to request accesss to the appropriate project vault.
A Dockerfile is provided in this repository to make it easier to run the scripts. A Makefile is provided with some very simple commands to build and run the image.
You can run make build update_assets
to build the Docker image and then generate the assets and database files.
Expect this process to take between 2 and 4 hours.
To build without Docker, there are some dependencies you will need to have available on your PATH.
These scripts are written in Python 3 and will not work with Python 2.
If you are using OSX you can install it via
brew install python3
Debian/Ubuntu users have access to the python3 command.
Image compression requires the mogrify
and convert
commands from imagemagick
If you are using OSX you can install it via
brew install imagemagick
Debian/Ubuntu users will need to install it with
sudo apt-get install imagemagick
Image compression also uses optipng
as the final step
If you are using OSX you can install it via
brew install optipng
Debian/Ubuntu users will need to install it with
sudo apt-get install optipng
cd /path/to/this/repo
python3 ./build-assets-from-signbank.py
Expect this process to take between 2 and 4 hours. A valid Signbank user account is required and may be configured via the following environment variables:
DEFAULT_SIGNBANK_HOST
SIGNBANK_USERNAME
SIGNBANK_PASSWORD
This is a holistic script that performs all the steps needed to get all the necessary assets from the NZSL Signbank server. It does a number of steps:
- Step 1: Fetching the latest signs from Signbank
- Step 2: Fetching images from Signbank
- Step 3: Rename all images to meet Android requirements
- Step 4: Write out nzsl.dat for Android
- Step 5: Write out sqlite nzsl.db for iOS
- Step 6: Merge images together into one folder
- Step 6a: Generate search thumbnails
- Step 6b: Shrink images for distribution
- Step 7: Cleanup (optional, requires -c flag)
At the end of this process, you will find pictures (diagrams + search thumbnails), in the 'assets/` folder, and the application databases in 'nzsl.dat' and 'nzsl.db'. The content of the two data files is the same, but they represent the data in different formats.
This script can be passed the following options:
--skip-assets
: The sign data will be extracted, but not assets. This substantially speeds up the export process, much of which is concerned with image resizing and transformation.--cleanup
: Remove the exported files as soon as the script completes--prerelease
: Export signs that are in the web-ready: check stage. This is determined by theSIGNBANK_WEB_READY_TAG_ID
environmnent variable.
Helper functions to interact with Signbank, DSRU's editorial system for managing the dictionary.
Helper functions to resize, compress and otherwise transform sign illustrations for app use.
Helper functions to interact with freelex and modify the files on the local file system. Deprecated and unused, this file is kept for historical reference only.
Prior extraction script for the previous NZSL editorial system, Freelex. Deprecated and unused, this file is kept for historical reference only.