A River for ElasticSearch to index subversion repositories.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Subversion River for ElasticSearch

Subversion River plugin for ElasticSearch

Getting Started


Use the elasticsearch plugin manager to install the plugin :

$ /path-to/elasticsearch/bin/plugin --install com.github.plombard/elasticsearch-river-subversion/0.3.5

Creating a Subversion river

Just create a new river of type "svn" and give it at least a repository and a path to index ("/" for the entire repos) :

curl -XPUT 'localhost:9200/_river/mysvnriver/_meta' -d '{
   "type": "svn",
   "svn": {
    "repos": "file:///myRepository",
    "path":  "mySvnModule/trunk"

River parameters

Except "repos" and "path", there are other parameters to the creation of the river :

"login", the user to use to connect to the repository (default "anonymous")

"password", the password to use to connect to the repository (default "password")

"update_rate", in ms, the time interval between every tick of the river (default 900000, so 15 mn)

"type", if you want your subversion information to use another type for the river (default "svn")

"bulk_size", the size of the bulks sent to the indexer (default 200)

"start_revision", in long format, the starting revision for the indexing (default "1L")

Data indexed

For the moment the river indexes only a few metadatas from svn, mainly the file content (I wrote it for full-text searches) : path, name, author, repository, revision, date, size, message and content.

Fields from and origin are respectively the revision and path of origin of the change, like when a tag is created from the trunk.

Please be aware that the mapping did change from the previous version, as the indexing of revisions is now preferred to the indexing of single documents.


Be very careful not to be too greedy with the bulk size and update rate parameters, as the retrieving of revisions can have a high memory consumption, and the indexing itself is very CPU-intensive.


These sources borrows heavily from :


I put most of the code under Apache 2.0 licence.