Permalink
Browse files

v0.1 release

* Added batch support, for much faster intiialization of current DB or reindexing all DB.
* Dropped indexes per model, instead, using `node_auto_index` and `relationship_auto_index`, letting Neo4j auto index objects.
* One `neo_save` method instead of `neo_create` and `neo_update`. It takes care of inserting or updating.
  • Loading branch information...
1 parent 2b1468b commit ead47a6d0d019f2341945d5bf044dc3a122ab43a @elado committed Jan 8, 2013
View
@@ -1,3 +1,38 @@
+## v0.1
+
+* Added batch support, for much faster intiialization of current DB or reindexing all DB.
+* Dropped indexes per model, instead, using `node_auto_index` and `relationship_auto_index`, letting Neo4j auto index objects.
+* One `neo_save` method instead of `neo_create` and `neo_update`. It takes care of inserting or updating.
+
+### Breaking changes:
+
+Model indexes (such as `users_index`) are now turned off by default. Instead, Neoid uses Neo4j's auto indexing feature.
+
+In order to have the model indexes back, use this in your configuration:
+
+```ruby
+Neoid.configure do |c|
+ c.enable_per_model_indexes = true
+end
+```
+
+This will turn on for all models.
+
+You can turn off for a specific model with:
+
+```ruby
+class User < ActiveRecord::Base
+ include Neoid::Node
+
+ neoidable enable_model_index: false do |c|
+ end
+end
+```
+
+## v0.0.51
+
+* Releasing Neoid as a gem.
+
## v0.0.41
* fixed really annoying bug caused by Rails design -- Rails doesn't call `after_destroy` when assigning many to many relationships to a model, like `user.movies = [m1, m2, m3]` or `user.update_attributes(params[:user])` where it contains `params[:user][:movie_ids]` list (say from checkboxes), but it DOES CALL after_create for the new relationships. the fix adds after_remove callback to the has_many relationships, ensuring neo4j is up to date with all changes, no matter how they were committed
View
@@ -3,7 +3,6 @@
[![Build Status](https://secure.travis-ci.org/elado/neoid.png)](http://travis-ci.org/elado/neoid)
-
Make your ActiveRecords stored and searchable on Neo4j graph database, in order to make fast graph queries that MySQL would crawl while doing them.
Neoid to Neo4j is like Sunspot to Solr. You get the benefits of Neo4j speed while keeping your schema on your plain old RDBMS.
@@ -12,18 +11,21 @@ Neoid doesn't require JRuby. It's based on the great [Neography](https://github.
Neoid offers querying Neo4j for IDs of objects and then fetch them from your RDBMS, or storing all desired data on Neo4j.
+**Important: Heroku Support is not available because Herokud doesn't support Gremlin. So until further notice, easiest way is to self host a Neo4j on EC2 in the same zone, and connect from your dyno to it**
+
+## Changelog
+
+[See Changelog](https://github.com/elado/neoid/blob/master/CHANGELOG.md)
## Installation
Add to your Gemfile and run the `bundle` command to install it.
```ruby
-gem 'neoid', '~> 0.0.51'
+gem 'neoid', '~> 0.1'
```
-Future versions may have breaking changes but will arrive with migration code.
-
**Requires Ruby 1.9.2 or later.**
## Usage
@@ -51,6 +53,11 @@ Neography.configure do |c|
end
Neoid.db = $neo
+
+Neoid.configure do |c|
+ # should Neoid create sub-reference from the ref node (id#0) to every node-model? default: true
+ c.enable_subrefs = true
+end
```
`01_` in the file name is in order to get this file loaded first, before the models (initializers are loaded alphabetically).
@@ -71,9 +78,9 @@ class User < ActiveRecord::Base
end
```
-This will help to create a corresponding node on Neo4j when a user is created, delete it when a user is destroyed, and update it if needed.
+This will help to create/update/destroy a corresponding node on Neo4j when changed are made a User model.
-Then, you can customize what fields will be saved on the node in Neo4j, inside neoidable configuration:
+Then, you can customize what fields will be saved on the node in Neo4j, inside `neoidable` configuration, using `field`. You can also pass blocks to save content that's not a real column:
```ruby
class User < ActiveRecord::Base
@@ -89,7 +96,6 @@ class User < ActiveRecord::Base
end
```
-
#### Relationships
Let's assume that a `User` can `Like` `Movie`s:
@@ -151,7 +157,7 @@ class Like < ActiveRecord::Base
end
```
-Neoid adds `neo_node` and `neo_relationships` to nodes and relationships, respectively.
+Neoid adds the metohds `neo_node` and `neo_relationships` to instances of nodes and relationships, respectively.
So you could do:
@@ -169,46 +175,60 @@ rel.end_node # user.movies.first.neo_node
rel.rel_type # 'likes'
```
-## Index for Full-Text Search
+#### Disabling auto saving to Neo4j:
-Using `search` block inside a `neoidable` block, you can store certain fields.
+If you'd like to save nodes manually rather than after_save, use `auto_index: false`:
```ruby
-# movie.rb
-
-class Movie < ActiveRecord::Base
+class User < ActiveRecord::Base
include Neoid::Node
-
- neoidable do |c|
- c.field :slug
- c.field :name
-
- c.search do |s|
- # full-text index fields
- s.fulltext :name
- s.fulltext :description
-
- # just index for exact matches
- s.index :year
- end
+
+ neoidable auto_index: false do |c|
end
end
-```
-Records will be automatically indexed when inserted or updated.
+user = User.create!(name: "Elad") # no node is created in Neo4j!
+
+user.neo_save # now there is!
+```
## Querying
You can query with all [Neography](https://github.com/maxdemarzi/neography)'s API: `traverse`, `execute_query` for Cypher, and `execute_script` for Gremlin.
+### Basics:
+
+#### Finding a node by ID
+
+Nodes and relationships are auto indexed in the `node_auto_index` and `relationship_auto_index` indexes, where the key is `Neoid::UNIQUE_ID_KEY` (which is 'neoid_unique_id') and the value is a combination of the class name and model id, `Movie:43`, this value is accessible with `model.neo_unique_id`. So use the constant and this method, never rely on assebling those values on your own because they might change in the future.
+
+That means, you can query like this:
+
+```ruby
+Neoid.db.get_node_auto_index(Neoid::UNIQUE_ID_KEY, user.neo_unique_id)
+# => returns a Neography hash
+
+Neoid::Node.from_hash(Neoid.db.get_node_auto_index(Neoid::UNIQUE_ID_KEY, user.neo_unique_id))
+# => returns a Neography::Node
+```
+
+#### Finding all nodes of type
+
+If Subreferences are enabled, you can get the subref node and then get all attached nodes:
+
+```ruby
+Neoid.ref_node.outgoing('users_subref').first.outgoing('users_subref').to_a
+# => this, according to Neography, returns an array of Neography::Node so no conversion is needed
+```
+
### Gremlin Example:
These examples query Neo4j using Gremlin for IDs of objects, and then fetches them from ActiveRecord with an `in` query.
Of course, you can store using the `neoidable do |c| c.field ... end` all the data you need in Neo4j and avoid querying ActiveRecord.
-**Most popular categories**
+**Most liked movies**
```ruby
gremlin_query = <<-GREMLIN
@@ -228,15 +248,18 @@ movie_ids = Neoid.db.execute_script(gremlin_query)
Movie.where(id: movie_ids)
```
-Assuming we have another `Friendship` model which is a relationship with start/end nodes of `user` and type of `friends`,
+*Side note: the resulted movies won't be sorted by like count because the RDBMS won't necessarily do it as we passed a list of IDs. You can sort it yourself with array manipulation, since you have the ids.*
+
**Movies of user friends that the user doesn't have**
+Let's assume we have another `Friendship` model which is a relationship with start/end nodes of `user` and type of `friends`,
+
```ruby
user = User.find(1)
gremlin_query = <<-GREMLIN
- u = g.idx('users_index')[[ar_id:user_id]].next()
+ u = g.idx('node_auto_index').get(unique_id_key, user_unique_id).next()
movies = []
u
@@ -246,15 +269,42 @@ gremlin_query = <<-GREMLIN
.except(movies).collect{it.ar_id}
GREMLIN
-movie_ids = Neoid.db.execute_script(gremlin_query, user_id: user.id)
+movie_ids = Neoid.db.execute_script(gremlin_query, unique_id_key: Neoid::UNIQUE_ID_KEY, user_unique_id: user.neo_unique_id)
Movie.where(id: movie_ids)
```
-`.next()` is in order to get a vertex object which we can actually query on.
+## Full Text Search
+### Index for Full-Text Search
-### Full Text Search
+Using `search` block inside a `neoidable` block, you can store certain fields.
+
+```ruby
+# movie.rb
+
+class Movie < ActiveRecord::Base
+ include Neoid::Node
+
+ neoidable do |c|
+ c.field :slug
+ c.field :name
+
+ c.search do |s|
+ # full-text index fields
+ s.fulltext :name
+ s.fulltext :description
+
+ # just index for exact matches
+ s.index :year
+ end
+ end
+end
+```
+
+Records will be automatically indexed when inserted or updated.
+
+### Querying a Full-Text Search index
```ruby
# will match all movies with full-text match for name/description. returns ActiveRecord instanced
@@ -270,14 +320,63 @@ Neoid.neo_search([Movie, User], "hello")
Movie.neo_search(year: 2013).results
```
+Full text search with Neoid is very limited and is likely not to develop more than this basic functionality. I strongly recommend using gems like Sunspot over Solr.
+
+## Batches
+
+Neoid has a batch ability, that is good for mass updateing/inserting of nodes/relationships. It sends batched requests to Neography, and takes care of type conversion (neography batch returns hashes and other primitive types) and "after" actions (via promises).
+
+A few examples, easy to complex:
+
+```ruby
+Neoid.batch(batch_size: 100) do
+ User.all.each(&:neo_save)
+end
+```
+With `then`:
+
+```ruby
+User.first.name # => "Elad"
+
+Neoid.batch(batch_size: 100) do
+ User.all.each(&:neo_save)
+end.then do |results|
+ # results is an array of the script results from neo4j REST.
+
+ results[0].name # => "Elad"
+end
+```
+
+*Nodes and relationships in the results are automatically converted to Neography::Node and Neography::Relationship, respectively.*
+
+With individual `then` as well as `then` for the entire batch:
+
+```ruby
+Neoid.batch(batch_size: 30) do |batch|
+ (1..90).each do |i|
+ (batch << [:create_node, { name: "Hello #{i}" }]).then { |result| puts result.name }
+ end
+end.then do |results|
+ puts results.collect(&:name)
+end
+```
+
+When in a batch, `neo_save` adds gremlin scripts to a batch, instead of running them immediately. The batch flushes whenever the `batch_size` option is met.
+So even if you have 20000 users, Neoid will insert/update in smaller batches. Default `batch_size` is 200.
+
+
## Inserting records of existing app
-If you have an existing database and just want to integrate Neoid, configure the `neoidable`s and run in a rake task or console
+If you have an existing database and just want to integrate Neoid, configure the `neoidable`s and run in a rake task or console.
+
+Use batches! It's free, and much faster. Also, you should use `includes` to incude the relationship edges on relationship entities, so it doesn't query the DB on each relationship.
```ruby
-[ Like.includes(:user).includes(:movie), OtherRelationshipModel ].each { |model| model.all.each(&:neo_update) }
+Neoid.batch do
+ [ Like.includes(:user).includes(:movie), OtherRelationshipModel.includes(:from_model).includes(:to_model) ].each { |model| model.all.each(&:neo_save) }
-NodeModel.all.each(&:neo_update)
+ NodeModel.all.each(&:neo_save)
+end
```
This will loop through all of your relationship records and generate the two edge nodes along with a relationship (eager loading for better performance).
@@ -289,30 +388,32 @@ Better interface for that in the future.
## Behind The Scenes
-Whenever the `neo_node` on nodes or `neo_relationship` on relationships is called, Neoid checks if there's a corresponding node/relationship in Neo4j. If not, it does the following:
+Whenever the `neo_node` on nodes or `neo_relationship` on relationships is called, Neoid checks if there's a corresponding node/relationship in Neo4j (with the auto indexes). If not, it does the following:
### For Nodes:
-1. Ensures there's a sub reference node (read [here](http://docs.neo4j.org/chunked/stable/tutorials-java-embedded-index.html) about sub reference nodes)
+1. Ensures there's a sub reference node (read [here](http://docs.neo4j.org/chunked/stable/tutorials-java-embedded-index.html) about sub references), if that option is on.
2. Creates a node based on the ActiveRecord, with the `id` attribute and all other attributes from `neoidable`'s field list
3. Creates a relationship between the sub reference node and the newly created node
-4. Adds the ActiveRecord `id` to a node index, pointing to the Neo4j node id, for fast lookup in the future
+4. Auto indexes a node in the auto index, for fast lookup in the future
-Then, when it needs to find it again, it just seeks the node index with that ActiveRecord id for its neo node id.
+Then, when it needs to find it again, it just seeks the auto index with that ActiveRecord id.
### For Relationships:
-Like Nodes, it uses an index (relationship index) to look up a relationship by ActiveRecord id
+Like Nodes, it uses an auto index, to look up a relationship by ActiveRecord id
1. With the options passed in the `neoidable`, it fetches the `start_node` and `end_node`
2. Then, it calls `neo_node` on both, in order to create the Neo4j nodes if they're not created yet, and creates the relationship with the type from the options.
-3. Add the relationship to the relationship index.
+3. Adds the relationship to the relationship index.
## Testing
In order to test your app or this gem, you need a running Neo4j database, dedicated to tests.
-I use port 7574 for this. To run another database locally:
+I use port 7574 for testing.
+
+To run another database locally (read [here](http://docs.neo4j.org/chunked/1.9.M03/server-installation.html#_multiple_server_instances_on_one_machine) too):
Copy the entire Neo4j database folder to a different location,
@@ -344,7 +445,7 @@ end
## Testing This Gem
-Just run `rake` from the gem folder.
+Run the Neo4j DB on port 7574, and run `rake` from the gem folder.
## Contributing
@@ -356,9 +457,9 @@ Please create a [new issue](https://github.com/elado/neoid/issues) if you run in
Unfortunately, as for now, Neo4j add-on on Heroku doesn't support Gremlin. Therefore, this gem won't work on Heroku's add on. You should self-host a Neo4j instance on an EC2 or any other server.
-## To Do
+## TO DO
-[To Do](https://github.com/elado/neoid/blob/master/TODO.md)
+[TO DO](HTTPS://GITHUB.COM/ELADO/NEOID/BLOB/MASTER/TODO.MD)
---
View
@@ -1,6 +1,4 @@
# Neoid - To Do
-* Allow to disable sub reference nodes through options
* Execute queries/scripts from model and not Neography (e.g. `Movie.neo_gremlin(gremlin_query)` with query that outputs IDs, returns a list of `Movie`s)
* Rake task to index all nodes and relatiohsips in Neo4j
-* Test update node
Oops, something went wrong.

0 comments on commit ead47a6

Please sign in to comment.