With the release of Doctrine MongoDB ODM 1.2, we’ve introduced several new features. I want to highlight some of them in this blog post and give an outlook on future development of MongoDB ODM.
With aggregation builder, we’re giving you a tool to create aggregation pipeline queries similar to how you create other queries in ODM. The builder takes into account mapping information to convert field names and types and also allows you to receive your results as mapped documents.
Previously, you had to run all aggregation pipeline queries through the collection object directly, receiving a list of arrays as result:
With the new aggregation builder, you can use a fluent builder interface and you don’t have to manually convert objects to their MongoDB types. To get result objects from an aggregation pipeline query, first create a
QueryResultDocument to hold the aggregated data:
You may note that the query result document does not have an identifier. Instead, we specify that the database name for our
user reference is
_id. This allows us to have a reference to a different object as identifier. The user reference will be returned as proxy object and only is initialized when first accessing any property of it.
Now you can create your aggregation pipeline query and use the
hydrate method to get objects as result:
The aggregation builder supports all aggregation pipeline stages present in MongoDB 3.4, including
$graphLookup. When looking up references, you can simply specify the name of the reference to look up and the pipeline stage will do the rest for you.
Due to a limitation in MongoDB,
$graphLookup can’t be used with references that are stored as
DBRef objects, only with references stored as
id or the new
DBRef replacement for storing references
As mentioned above, MongoDB does not support looking up references from
DBRef objects. This is due to the field names in
DBRef which contain a leading dollar sign. Storing your references as
id works, but makes them incompatible with discriminators. To solve this problem, we introduced the new
ref storage type for references. It works like a
DBRef object, but only stores the identifier of the referenced object and any discriminator field if present in the mapping. This new type is opt-in, so as long as you’re not using references in aggregation pipelines, you can happily keep using
DBRef for storing your references. To use this new storage type, simply specify a value of
storeAs in your reference mapping:
Using primers for inverse references
Reference priming helps you avoid n+1 problems by initializing proxy objects of the same type at once, resulting in one query instead of one query for each proxy. You could enable reference priming in the query builder:
In inverse references, the only way to prime other references was to map the inverse reference using a
repositoryMethod and build the query manually. To avoid this code overhead, you can now specify which fields to prime in an inverse reference with
Note that for inverse references mapped with
repositoryMethod, the primers in the mapping take precedence and override any primers that might have been set on the query already.
With 1.2.0 released, development now focuses on replacing the legacy MongoDB driver (
ext-mongo) with its new counterpart,
ext-mongodb. This also allows you to run ODM on PHP 7+ without the need of polyfill libraries like
mongo-php-adapter. We hope to release a first beta version of ODM 2.0 before the end of the year, with a final release no later than the first quarter of 2018 (no, we’re not going to do another 5-year beta phase).
We’ve already started deprecating features like
requireIndexes and much more in 1.2 and will make several more deprecations as we work on 2.0. Since deprecations can only be introduced in minor releases, we will release 1.3.0 at about the same time as the first beta for ODM 2.0. This version will deprecate all features to be removed in 2.0, and we’ll also try to add a forward compatibility layer to 2.0 where possible. It will also be the last release compatible with PHP 5.6, and we’ll continue to support it for 6 months after the release of ODM 2.0.