fix: prod / cr-lf in metadata is ok for aggregate filter
cr,lf,crlf are normalized
- getting field values from recordadapter
- during es indexation (direct sql read in metadata...)
- before querying
Here is the new model:
+-----------------------------+
| Structure |
+-----------------------------+
| +createFromDataboxes() |
| getAllFields() |
| getUnrestrictedFields() |
| getPrivateFields() |
| getFacetsFields() |
| getThesaurusEnabledFields() |
| getDateFields() |
|- - - - - - - - - - - - - - -|
| add() |
| get() |
| typeOf() |
| isPrivate() |
+-------+-+-+-----------------+
| | | +---------------------+
| | +--------> | Field |
| | +---------------------+
| | | getName() |
| | | getType() |
| | | isXXX() |
| | | getThesaurusRoots() |
| | +---------------------+
| |
| | +-------+
| +----------> | Field |
| +-------+
|
| +-------+
+------------> | Field |
+-------+
It was driven by the following use cases:
- Get list of facets (only searchable fields)
- Get list of fields with concept inference
- Get list of all fields
- Splitted in private / public fields (to define mapping)
- Get all date fields
- Get field type
- To apply sanitization rules
- To define mapping
- Check if concept inference enabled
- Check if the field is searchable
- Check if the field is a facet
- Check if the field is private
- Dereference field from label (still to be done)
(The last two UCs are new)
Also removed old code from legacy search engines.
[#PHRAS-500]
- Date and number types sanitization
- Remove `RecordIndexer` dependency on `ElasticSearchEngine`
- Move some sanitization from `RecordIndexer` to `RecordHelper`
Clicking on a facet value on the left pane now return the expected result count.
This commit implement a new "raw" matcher. It can be used like
`r"some raw value"`. It operate on the the `.raw` multi-field and skips all
analysis.
Escaping `"` is supported by prepending a backslash `\"`. You can also escape
the escaping character `\` by doubling it (`\\`).
Adds a new `ContextAbleInterface` to differenciate matcher supporting an
optional context from those who can't.
Fixes an issue with `QueryContext::narrowToFields()` ignoring passed fields.
- Index the full content of a record in a (private_)content_all field
- Handle all fields wide search as a special-case (drastically simplify queries)
- QueryContext doesn't take all allowed fields anymore, but whether private
fields are allowed or not. Since private fields are namespaced, field level
restriction is not needed anymore.
