tree: c84bbc21da9f7506e0dca3a9f08a080a28997622 [path history] [tgz]

go-git + aerospike: a git repository backed by a database

This is an example of a go-git repository backed by Aerospike.

and what this means ...

git has as very well defined storage system, the .git directory, present on any repository. This is the place where git stores al the objects, references and configuration. This information is stored in plain files.

Our original go-git version was designed to work in memory, some time after we added support to read the .git, and now we have added support for fully customized storages.

This means that the internal database of any repository can be saved and accessed on any support, databases, distributed filesystems, etc. This functionality is pretty similar to the libgit2 backends


What do you need? You need an aerospike server. The easiest way to get one for testing is running the official docker container provided by Aerospike:

docker run -d -p 3000:3000 --name aerospike aerospike/aerospike-server

Now, we need the sample code.

go get -u

Running this command will make the binary go-git-aerospike. if you have GOPATH on your PATH, you are ready to go. If not, this is a great moment.


Cloning the repository into the database

Running the command go-git-aerospike with the clone option followed by the URL of a git repository clones the repository into the database, storing all the git objects in it:

go-git-aerospike clone

The repository is stored in the aerospike database. This means that all the internal objects like commits, trees, blobs and tags are records in different sets in the test namespace:

aql> SELECT hash, type, url FROM test.commit
| hash                                       | type     | blob  | url                               |
| "c94450c805876e49b38d2ff1103b8c09cdd2aef4" | "commit" | 00 00 |" |
| "7f71640877608ee9cfe584fac216f03f9aebb523" | "commit" | 00 00 |" |
| "255f097450dd91812c4eb7b9e0d3a4f034f2acaf" | "commit" | 00 00 |" |
102 rows in set (0.071 secs)

And also the references and the configuration (remotes) are stored in it.

aql> SELECT name, target, url FROM test.reference
| name                         | target                                     | url                   |
| "HEAD"                       | "ref: refs/heads/master"                   | ...rc-d/flamingo.git" |
| "refs/heads/master"          | "ed3e1aa2e46584cb803ed356cb5d8855f6d05660" | ...rc-d/flamingo.git" |
| "refs/remotes/origin/master" | "ed3e1aa2e46584cb803ed356cb5d8855f6d05660" | ...rc-d/flamingo.git" |
3 rows in set (0.046 secs)

Reading the repository

Running the log command, a git log --online like result is printed:

go-git-aerospike log

The URL of the repository is the way we identify the objects in the sets, since we can clone several repositories to the same database.

ed3e1aa ID is also allowed in SendFormTo and SayTo
2031f3e Handle close of message channel in WaitForMessage
e784495 Add SendFormTo and SayTo
447748a Make text in attachments accept markdown
595b4e7 Form author name and author icon and text groupfield
0f2e315 Test for InvokeAction
0dc7c9a Handle closing of channel
b3f167b Implement InvokeAction

The process has read all the commits and all the needed objects from the aerospike sets.

Playing with the database

If you want to explore the database, you can execute the aql tool and run some queries:

docker run -it aerospike/aerospike-tools aql -h
aql> SELECT * FROM test;