Lookup syntax

Many tasks require users to specify artifacts or collections in some way. Rather than having people write IDs everywhere, it’s better to have a query language with a more expressive syntax.

This syntax has four forms with different YAML types:

String lookup

A string lookup consists of one or more segments, separated by the slash character (/).

The first segment identifies a collection as name@category (e.g. debian@debian:archive). Some lookup contexts may provide a default category, in which case the first segment may just be a name.

Each subsequent segment is looked up using the lookup method on the appropriate collection manager for the collection returned by traversing up to the previous segment. Segments that do not contain a colon are syntactic sugar for the name: lookup (that is, the segment abc is looked up as name:abc). It is an error to specify more segments after one that does not return a collection.

Some strings have a special meaning:

  • NNN@artifacts or NNN@collections looks up an artifact or collection by ID.

  • In the context of a parent workflow, internal@collections is syntactic sugar for that workflow’s internal collection; it is equivalent to workflow-NNN@debusine:workflow-internal, where NNN is the ID of the relevant workflow work request (with task_type: workflow).

This lookup always returns exactly one artifact or collection. It is an error if no match is found.

Integer lookup

An integer lookup looks up an artifact or collection by ID. For this lookup type to be valid, the context must specify that it expects an artifact or a collection respectively.

Single lookup

A lookup for a single item is a String lookup or an Integer lookup.

Dictionary lookup

A dictionary lookup specifies a collection and a list of filters that are applied to it, returning all the artifacts or collections that match.

The collection key provides a string lookup (as above) which is resolved to a collection. It is an error if this lookup returns a kind of item other than a collection. If this lookup is well-formed but returns nothing, then the dictionary lookup returns no items.

All other keys are filters on properties of the active items in this collection. Only certain keys are allowed:

  • child_type: match items with the given type (bare, artifact, or collection), defaulting to artifact. (Note to implementers: the choices here are not quite the same as the internal database representation.)

  • category: match items with the given category.

  • name: match items with the given name.

  • data__KEY: match items whose per-item data has a field called KEY with the given value. KEY may not contain __.

In addition, some Django-style suffixes may be added to name or data__KEY to request different matching strategies, as follows:

  • __contains: the value contains the given substring

  • __endswith: the value ends with the given string

  • __startswith: the value starts with the given string

Other __ suffixes are not currently supported; nor is using __ to navigate deeper structures embedded in per-item data, or using __ suffixes on keys other than name or data__KEY.

For example, the following lookup returns all the active debian:binary-package artifacts in the collection looked up by debian/trixie that have package: "libc6" and version: "2.37-15" in their per-item data:

collection: debian/trixie
child_type: artifact
category: debian:binary-package
data__package: libc6
data__version: "2.37-15"

List lookup

A list lookup returns all the artifacts or collections that match any one of the items in the list, where each item is a string, integer, or dictionary lookup as above.

For example, the following lookup returns all the active items in the collection looked up by debian/trixie that have either (1) package: "libc6-dev" and version: "2.37-15" or (2) have names starting with debhelper_13.15.3_, and also the artifact with ID 123:

- collection: debian/trixie
  data__package: libc6-dev
  data__version: "2.37-15"
- collection: debian/trixie
  name__startswith: debhelper_13.15.3_
- 123@artifacts

Multiple lookup

A lookup for multiple items is a Dictionary lookup or a List lookup.