Cannot find route to use

[mlovrovich]

    const http = require('http')
    const router = require('find-my-way')()

    router.on('GET', '/:namespace/:type/:id', (req, res, params) => {
        res.end('{"message":"hello world"}')
    })

    router.on('POST', '/:namespace/jobs/:name/run', (req, res, params) => {
        res.end('{"message":"hello world"}')
    })

    const server = http.createServer((req, res) => {
        router.lookup(req, res)
    })

    server.listen(4000, err => {
        if (err) throw err
        console.log('Server listening on: http://localhost:4000')
    })

Works as expected

GET http://localhost:4000/test_namespace/test_type/test_id
GET http://localhost:4000/test_namespace/jobss/test_id

Does not work

GET http://localhost:4000/test_namespace/jobs/test_id

[delvedor]

Hello!
Can you retry with the latest release?

[mlovrovich]

@delvedor hello, the same issue is happening with the latest release 1.17.1

[mcollina]

I think the only way to fix this us to have a different trie for every http method.
It’s a big refactoring, and I fear neither Tomas or myself have time now.

[IpaliboWhyte]

Another way to fix this issue could be storing some useful information on each trie node and use those information to make decision in Node.prototype.findChild and Node.prototype.findVersionChild

[mcollina]

@IpaliboWhyte what would you have in mind?

[IpaliboWhyte]

@mcollina Hey, sorry for the late reply. The problem I see so far is that we go too far in the tree. In this case we go too far to search for test_namespace/jobs/test_id in the node for route: /test_namespace/jobs/:name/run.

This issue is also reproducible with these two routes
/jobs/:name/run
/:type/:id

and this search query => jobs/test_id

In theory, if nodes have some stored property which describes the subpath length of a route (based on slash delimiters), by using the search query subpath length we could do a best fit scan to return a more accurate child node. My guess is that findChild will calls this first before proceeding.

Example:

=> /:type/:id (length of 2)
=> /jobs/:name/run (length of 3)

The path /jobs/test_id would have a length of 2, in that case we return the child node /:type/:id. It will be useless to go to /jobs/:name/run as it has a length of 3.

Just my thought for now, hopefully this makes sense :)

[delvedor]

Not sure that the approach you describe will work.
We are using a radix tree, which means that a Node can contain multiple slashes.
Take the following example:

Route 1: /hello/work
Route 2: /hello/wonderful/home/kitchen
Route 3: /heaven/cloud

The generated tree will be

/he
  |-aven/cloud
  |-llo/wo
     |-rk
     |-nderful/home/kitchen

If we start splitting on the slash when a new request comes in we will lose all the benefits of the above data structure, and that point it will be a better idea refactor the entire module (already did it, the performances were not as good).

[IpaliboWhyte]

good point, might affect performance alot. I'll try to take a deep dive into this when I get some time :)

[ivan-tymoshenko]

This issue was fixed. I added a test here #261. We can close this.