Query Git repositories using GraphQL, and explore the source code using Babelfish Universal Abstract Syntax Tree (UAST). Built on top of gitbase.
This project is a proof of concept.
First you will need a directory containing some Git repositories:
mkdir $HOME/repos cd $HOME/repos git clone https://github.com/src-d/gitbase.git git clone https://github.com/src-d/gitbase-web.git git clone https://github.com/src-d/engine.gitThen use docker-compose to automatically run gitbase-graphql, gitbase and bblfshd.
wget https://raw.githubusercontent.com/carlosms/gitbase-graphql/master/docker-compose.yml docker-compose pull GITBASE_GQL_REPOS_FOLDER=$HOME/repos docker-compose up --force-recreateNow go to http://localhost:3000/graphiql and explore!
In the following link you will find the GraphQL schema definition.
Some queries you can try:
{ allRepositories { id refs { name isTag isRemote } remotes{ name fetchUrl } } }{ allRepositories { id refs(name: "HEAD") { name commits { authorName message hash } } } }{ allRepositories { id refs(name: "HEAD") { commit { jsFiles: files(language: "JavaScript") { path size } goFiles: files(language: "Go") { path size } } } } }You can use the raw functionality offered by Babelfish using uastRaw. Continue to the Babelfish documentation to learn about the available xpath querying.
{ repository(id: "gitbase-web") { id refs(name: "HEAD") { commit { treeEntries(name: "App.js") { name blob { uastRaw(language: "JavaScript") } } } } } }{ repository(id: "gitbase-web") { id refs(name: "HEAD") { commit { treeEntries(name: "App.js") { name blob { uastRaw(language: "JavaScript", xpath:"//*[@roleNumber and @roleLiteral]") } } } } } }Using uast you can query each tree node with GraphQL, selecting which fields to include in the response.
Note: each children level has to be queried explicitly. Although you cannot query all the tree nodes recursively, you can use childrenRaw to filter some nodes with GraphQL, and retrieving the complete tree dangling from that node.
{ repository(id: "gitbase-web") { id refs(name: "HEAD") { commit { treeEntries(language: "JavaScript") { name blob { uast(language: "JavaScript") { internal_type token roles children { internal_type token roles start_position { offset line col } end_position { offset line col } children { internal_type token children { internal_type token } } } } } } } } } }The fields uast and children accept filter arguments, for now limited to exact string matching on internal_type or token.
Instead of setting a filter for each level of the query, you can use the flat argument. This will flatten the tree, so you filter nodes ignoring the tree level they are at.
# All the imports done in App.js { repository(id: "gitbase-web") { id refs(name: "HEAD") { commit { treeEntries(name: "App.js") { name blob { uast(language: "JavaScript", flat: true, internal_type: "ImportDeclaration") { children(internal_type: "StringLiteral") { token } } } } } } } }{ repository(id: "gitbase-web") { id refs(name: "HEAD") { commit { treeEntries(name: "App.js") { name blob { uast(language: "JavaScript", flat: true, internal_type: "CallExpression") { children(flat: true, internal_type: "Identifier") { token start_position { offset line col } end_position { offset line col } } } } } } } } }There is an experimental argument o pass a sort of UDF (User Defined Function) to filter the nodes with any arbitrary code. The JavaScript code must be set as a string for filter_func. node and result are accesible as global variables. For example,
filter_func: "result = node.token.length > 15;"filter_func: "result = node.internal_type !== 'Comment' && node.token.length > 0;"
{ allRepositories { id refs(name: "HEAD") { commit { files(language: "Go") { path size uast(language: "Go", flat: true, filter_func: "result = node.token.length > 15;") { token internal_type } } } } } }To run the project from sources execute:
yarn install yarn startThe following environment variables can be set:
| Variable | Default | Description |
|---|---|---|
GITBASE_GQL_PORT | 3000 | Port where the GraphQL server server will listen |
GITBASE_GQL_DB_HOST | localhost | Host where the gitbase server is listening |
GITBASE_GQL_DB_PORT | 3306 | Port where the gitbase server is listening |
GITBASE_GQL_DB_USER | root | User name used for the gitbase server connection |
GITBASE_GQL_DB_PASSWORD | | Password used for the gitbase server connection |