Compilation time for routes with complex body definitions is non-linear

[hypersurge]

Compilation time (observed on JS and PHP) takes a big hit when creating a route with a body with many members (30 seconds or more for 50 members).

Here's a minimal example:

import tink.http.containers.*;
import tink.http.Response;
import tink.web.routing.*;

class Main
{
    static function main()
    {
        var l_container = PhpContainer.inst;
        var l_router = new Router<Root>( new Root() );
        l_container.run( p_request -> l_router.route( Context.ofRequest( p_request ) ).recover( OutgoingResponse.reportError ) );
    }
}

class Root
{
    public function new()
    {
    }    

    @:get( 'test' )
    @:produces( 'text/html' )
    public function getTest( body:
        {
            ?a1:String, ?a2:String, ?a3:String, ?a4:String, ?a5:String, ?a6:String, ?a7:String, ?a8:String, ?a9:String, ?a10:String,
            ?b1:String, ?b2:String, ?b3:String, ?b4:String, ?b5:String, ?b6:String, ?b7:String, ?b8:String, ?b9:String, ?b10:String,
            ?c1:String, ?c2:String, ?c3:String, ?c4:String, ?c5:String, ?c6:String, ?c7:String, ?c8:String, ?c9:String, ?c10:String,
            ?d1:String, ?d2:String, ?d3:String, ?d4:String, ?d5:String, ?d6:String, ?d7:String, ?d8:String, ?d9:String, ?d10:String,
            ?e1:String, ?e2:String, ?e3:String, ?e4:String, ?e5:String, ?e6:String, ?e7:String, ?e8:String, ?e9:String, ?e10:String,
        } 
        )
    {
        return "OK";
    }
}

Comment out all the "e"s and compile time is ~halved, comment out the "d"s and compile time is ~halved again, comment out the "c"s and compile time is negligible.

@back2dos described this on Gitter as "seems like the macro has non-linear complexity"

[hypersurge]

For others encountering this issue ... you may find an immediate workaround is to reduce the total number of unique body parameters (e.g. by using arrays or more complex objects).

[kevinresol]

Perhaps the non linear complexity comes from the conflict checker https://github.com/haxetink/tink_web/blob/3b2d1560c91966ce6b72aeb98d3aa6235e2eb183/src/tink/web/macros/Parameters.hx#L78-L106

[back2dos]

Ok, looks like the macro isn't the problem (I rewrote it with maps to avoid the double loop to no effect). Compiling with -D times gives me filters | 27.283 | 77 | 77 | 11 |, so something's wrong there. Tried to upgrade to nightly, but got stuck on some final variance issue. Will check later.

[back2dos]

Looking at the generated code, it seems like we're generating a variable called body 2500 times (there's 50 fields and 50 * 50 happens to be 2500).