LSP can't autocomplete for provider "google"

[GuillaumeDesforges]

Using the template provided with the TF provider github has autocomplete working.
But with TF provider google autocomplete does not work.

{
  inputs = {
    nixpkgs.url = "nixpkgs/nixos-unstable";
    nickel.url = "github:tweag/nickel";
    tf-ncl.url = "github:tweag/tf-ncl";
    tf-ncl.inputs.nickel.follows = "nickel";
    utils.url = "github:numtide/flake-utils";
  };

  nixConfig = {
    extra-substituters = [ "https://tweag-nickel.cachix.org" ];
    extra-trusted-public-keys = [ "tweag-nickel.cachix.org-1:GIthuiK4LRgnW64ALYEoioVUQBWs0jexyoYVeLDBwRA=" ];
  };

  outputs = inputs: inputs.utils.lib.eachDefaultSystem (system:
    {
      devShell = inputs.tf-ncl.lib.${system}.mkDevShell {
        providers = p: {
          inherit (p) github;
        };
      };
    });
}

{
  inputs = {
    nixpkgs.url = "nixpkgs/nixos-unstable";
    nickel.url = "github:tweag/nickel";
    tf-ncl.url = "github:tweag/tf-ncl";
    tf-ncl.inputs.nickel.follows = "nickel";
    utils.url = "github:numtide/flake-utils";
  };

  nixConfig = {
    extra-substituters = [ "https://tweag-nickel.cachix.org" ];
    extra-trusted-public-keys = [ "tweag-nickel.cachix.org-1:GIthuiK4LRgnW64ALYEoioVUQBWs0jexyoYVeLDBwRA=" ];
  };

  outputs = inputs: inputs.utils.lib.eachDefaultSystem (system:
    {
      devShell = inputs.tf-ncl.lib.${system}.mkDevShell {
        providers = p: {
          inherit (p) google;
        };
      };
    });
}

Note: don't forget to run link-schema in between

[vkleen]

Thanks for the detailed report! I'm fairly confident that the issue is the language server just taking too long to process the completion request for the google provider contract, on account of its size. The best solution to that would be a project that I was contemplating anyway, namely to try modularizing the generated contracts. It might be feasible to replace the current monolithic contract file with one that essentially just lists all the available resources or data source with import statements for the actualy contract for each. That would at least push the timeout limit further down the road.

[GuillaumeDesforges]

@vkleen thanks for the fast reply

the issue is the language server just taking too long to process the completion request for the google provider contract, on account of its size

I came to the same conclusions as well. For the record, I use coc.nvim.

modularizing the generated contracts

Sounds very interesting. 🙂

[sedlund]

tf-ncl-2024-03-19_16-48-11.webm

i tried this example with the github provider and found that the nls language server process climbed up to 5gb of used ram while trying different completions and it oom'd my VM i was running it on. is the ram usage expected? if so how much ram should i plan to have to work with this?

[yannham]

Hi @sedlund, thanks for the report. 5GB isn't reasonable, no. Could you tell what version of nls did you install? Because of a dumb overlook, nls doesn't have a --version argument available, but if you installed it with a nickel binary, then nickel --version should do the trick.

[sedlund]

didnt want to hijack this - opened #66

[yannham]

@jneem I believe this has been fixed by tweag/nickel#2073?

[jneem]

I believe so, yes.

[yannham]

Great, let's close - feel free to ask to re-open if this isn't the case in your experience.