KPI

Python Dependencies

Python dependencies are managed with pip-compile and pip-sync from the pip-tools package. The dependencies are listed in dependencies/pip/, with core requirements in dependencies/pip/requirements.in. You may use pip directly with one of the compiled dependencies/pip/*.txt files, but consider using instead the pip-sync. Do not add new dependencies directly to the compiled dependencies/pip/*.txt files; instead, update the relevant the source dependencies/pip/*.in file(s), and execute make pip_compile after any changes. You can pass arguments to pip-compile with e.g. make pip_compile ARGS='--upgrade-package=xlwt'. To force building, use make --always-make ....

Ubuntu 16.04 apt Dependencies

apt dependencies for Ubuntu 16.04 are listed in dependencies/apt_requirements.txt and can be installed with e.g. apt-get install $(cat dependencies/apt_requirements.txt).

Downloading and compiling the translations

• Pull the submodule into the locale directory with git submodule update.
• Optionally configure transifex to pull the latest translations into the locale directory with tx pull --all
• Run python manage.py compilemessages to create .mo files from the .po files.
• To test out locales in the interface, double click "account actions" in the left navbar, use the dropdown to select a language, and refresh.

Searching assets and collections

Top-level (null-parent) assets and collections can be found by including parent= in the query string. For other searches, construct a string using the Whoosh query language and pass it in as the q parameter, e.g. /assets/?q=name:sanitation. Fields indexed by Whoosh are:

• name: a tokenized* representation of the name;
• name__exact: a space- and comma-escaped representation of the name, e.g. "Fun, Exciting Asset" would be indexed as "Fun--Exciting-Asset";
• owner__username: a tokenized* representation of the owner's username;
• owner__username__exact: a space- and comma-escaped representation of the owner's username;
• parent__name: a tokenized* representation of the parent object's name;
• parent__name__exact: a space- and comma-escaped representation of the parent object's name;
• parent__uid: the UID of the parent collection;
• ancestor__uid: a multi-value field containing the UIDs of all ancestor collections;
• tag: a multi-valued field holding space- and comma-escaped representations of each tag assigned to the object;
• asset_type (for assets only): a space- and comma-escaped representation of the asset's type string;
• text: the search "document," which is built by text templates.

When the q parameter contains a search term without a specified field, e.g. /collections/?q=health, that term is matched against the search "document" (the text field).

Searching tags

Construct a string using the Whoosh query language and pass it in as the q parameter, e.g. /tags/?q=asset_type:block. Fields indexed by Whoosh are:

• name__ngram: the tag's name decomposed into n-grams, e.g. ?q=name__ngram:cat would match tags named "dogs/cats" and "education";
• asset_type: a multi-value field containing the types (e.g. form, question, block) of all tagged assets;
• kind: a multi-value field containing "asset" when assets are tagged, "collection" when collections are tagged, or both;
• text: a tokenized* representation of the name, which serves as the search "document" (see note below).

When the q parameter contains a search term without a specified field, e.g. /tags/?q=health, that term is matched against the search "document" (the text field).

* Implemented by Haystack as a Whoosh TEXT field using the StemmingAnalyzer. Unsuitable for exact matching.