Skip to content

Latest commit

 

History

History
69 lines (48 loc) · 2.61 KB

UPDGRADE-2.0-dev.md

File metadata and controls

69 lines (48 loc) · 2.61 KB
Raw

UPGRADE FROM 1.0-dev to 2.0-dev

Configuration of Elastica Index

In 1.0-dev version :

  • To sort on a field, it needs to be not_analyzed.
  • To filter on a field, it needs to be analyzed.

So to make both works, in the 1.0-dev version, all the fields (except id) must be set to multi_field with a raw sub-field: the raw sub-field is not_analyzed and is used to sort.

For more information about this, see ElasticSearch documentation (or this one, as the 1.0 version of ElasticSearch was released recently).

So the bundle won't work if one of your field does not have a raw sub-field. Which is not a good thing, especially as you don't need an analyzed field to do an exact match:

Example

If I have a status field that can be either "Not Started", "Started", "Finished".

  • To sort it, it need to be not_analyzed.
  • To filter it, it does not need to be analyzed if the filter is on the exact word (like when we sort from a select list).

In 2.0-dev version :

Configure fields as multi_field is not required anymore, so the sort is not done on the $fieldName .'.raw' anymore. If you upgrade the bundle, you will need to make some change in your index configuration.

For the fields that just need to be not_analyzed to perform "exact search" & filter, no need to use the multi_field.

For string fields, if you want to be able to sort and to search on the field, you may want to declare them as multi_field, with two "sub-fields"

  • The first one named after the field, required for search, must use index: analyzed
  • The second one named raw, required for sorting, must use index: not_analyzed

Example

book:
    mappings:
        id: {type: integer}
        title:
            type: multi_field
            fields:
                title: { type: string, index: analyzed }
                raw: { type: string, index: not_analyzed }
        created_at: { type: date }
        ...

Then in your Admin class, configure the field to use the not_analyzed sub-field to sort:

    protected function configureListFields(ListMapper $listMapper)
    {
        $listMapper
            ->add('title', 'string', array(
                'sortable' => true,
                'sort_field_mapping' => ["fieldName" => "title.raw", "type"=> "string"] // To be able to sort by title.raw which is not_analyzed
            ))
            ...
        ;
    }