Router radixtree

what's libradixtree?

libradixtree, adaptive radix trees implemented in Lua for OpenResty.

APISIX using libradixtree as route dispatching library.

How to use libradixtree in APISIX?

This is Lua-Openresty implementation library base on FFI for rax.

Let's take a look at a few examples and have an intuitive understanding.

1. Full match

/blog/foo

It will only match /blog/foo.

2. Prefix matching

/blog/bar*

It will match the path with the prefix /blog/bar, eg: /blog/bar/a, /blog/bar/b, /blog/bar/c/d/e, /blog/bar etc.

3. Match priority

Full match -> Deep prefix matching.

Here are the rules:

/blog/foo/*
/blog/foo/a/*
/blog/foo/c/*
/blog/foo/bar

path	Match result
/blog/foo/bar	/blog/foo/bar
/blog/foo/a/b/c	/blog/foo/a/*
/blog/foo/c/d	/blog/foo/c/*
/blog/foo/gloo	/blog/foo/*
/blog/bar	not match

4. Parameter match

When radixtree_uri_with_parameter is used, we can match routes with parameters.

For example, with configuration:

apisix:
    router:
        http: 'radixtree_uri_with_parameter'

route like

/blog/:name

will match both /blog/dog and /blog/cat.

For more details, see https://github.com/api7/lua-resty-radixtree/#parameters-in-path.

How to filter route by Nginx builtin variable

Please take a look at radixtree-new, here is an simple example:

$ curl http://127.0.0.1:9080/apisix/admin/routes/1 -H 'X-API-KEY: edd1c9f034335f136f87ad84b625c8f1' -X PUT -i -d '
{
    "uri": "/index.html",
    "vars": [
        ["http_host", "==", "iresty.com"],
        ["cookie_device_id", "==", "a66f0cdc4ba2df8c096f74c9110163a9"],
        ["arg_name", "==", "json"],
        ["arg_age", ">", "18"],
        ["arg_address", "~~", "China.*"]
    ],
    "upstream": {
        "type": "roundrobin",
        "nodes": {
            "39.97.63.215:80": 1
        }
    }
}'

This route will require the request header host equal iresty.com, request cookie key _device_id equal a66f0cdc4ba2df8c096f74c9110163a9 etc.

How to filter route by graphql attributes

APISIX supports filtering route by some attributes of graphql. Currently we support:

• graphql_operation
• graphql_name
• graphql_root_fields

For instance, with graphql like this:

query getRepo {
    owner {
        name
    }
    repo {
        created
    }
}

• The graphql_operation is query
• The graphql_name is getRepo,
• The graphql_root_fields is ["owner", "repo"]

We can filter such route out with:

$ curl http://127.0.0.1:9080/apisix/admin/routes/1 -H 'X-API-KEY: edd1c9f034335f136f87ad84b625c8f1' -X PUT -i -d '
{
    "methods": ["POST"],
    "uri": "/_graphql",
    "vars": [
        ["graphql_operation", "==", "query"],
        ["graphql_name", "==", "getRepo"],
        ["graphql_root_fields", "has", "owner"]
    ],
    "upstream": {
        "type": "roundrobin",
        "nodes": {
            "39.97.63.215:80": 1
        }
    }
}'

To prevent spending too much time reading invalid graphql request body, we only read the first 1 MiB data from the request body. This limitation is configured via:

graphql:
  max_size: 1048576

If you need to pass a graphql body which is larger than the limitation, you can increase the value in conf/config.yaml.