NLU Demo Project

There are 3 projects in this repo:

1. rolo
2. cholo
3. yolo

Each project depends on the 2 other projects, to demonstrate NLU working with circular deps. NLU works by searching your fs for relevant projects. To restrict the search you use an env variable, otherwise, NLU will search starting from $HOME, which is slower. So it's better to restrict it, which is easy to do.

Depending on what your main/primary project is, the tree looks like this:

      └─ cholo1
           ├─ yolo3
           │  ├─ cholo1
           │  └─ rolo2
           └─ rolo2
              ├─ cholo1
              └─ yolo3

or like this:

        └─ rolo2
           ├─ cholo1
           │  ├─ yolo3
           │  └─ rolo2
           └─ yolo3
              ├─ cholo1
              └─ rolo2

or like this:

        └─ yolo3
           ├─ cholo1
           │  ├─ yolo3
           │  └─ rolo2
           └─ rolo2
              ├─ cholo1
              └─ yolo3

=> (Your main/primary project is whatever project is in in your PWD/CWD).

Instructions on how to replicate/run the demo

1. Clone rolo-cholo-yolo on your local system

git clone https://github.com/ORESoftware/rolo-cholo-yolo.git

2. For each project (rolo, cholo, and yolo) run:

nlu init

When it prompts you for a searchRoot, you can enter this line:

$HOME/<location>/yolo:$HOME/<location>/rolo:$HOME/<location>/cholo

where <location> is the path to where you cloned the repo. Using environment varables for searchRoots allows you to work on projects with other people, or yourself, if you work on more than one machine.

So as another step, create this env var and add it to bashrc or bash_profile:

export my_nlu_search_root="$HOME/<location>";

Then in your .nlu.json files, you can use:

{
  "searchRoots":[
    "$my_nlu_search_root"
  ]
}

And for each developer/machine that works on rolo-yolo-cholo, they can define $my_nlu_search_root to whatever path they wish. Essentially each machine can have it's own definition for the searchRoot(s) of your dependency tree.

So, by the way, you have should have run something like this:

( cd rolo && nlu init )
( cd yolo && nlu init )
( cd cholo && nlu init )

3. Now run the link!

Now, in each .nlu.json file, there should be a array property called "list". Notice that "async", "lodash" and "ramda" did not make it into list, but "rolo2", "cholo1" and "yolo3" did. That is because NLU did not find local copies of ramda and lodash, but did find local copies of the other 3.

Now, you can cd into any project and run:

$ nlu run

It will link up the project with the other 2 projects. Trying running $ nlu run in all 3 projects. Visually inspect the node_modules folder for each project, so see the result.

You can use --dry or --dry-run

For reference, you can use nlu run --dry to get a visual of how the linkage will happen, without actually doing any writes.

Final note

You may be thinking - rolo-yolo-cholo is a monorepo. It sure is. But the reason why mono-repos are not the answer for simple development, is because some projects do not belong in the mono-repo but still need to be included as a dependency of projects in the monorepo, and so the question is then - how do we link projects that don't belong in the monorepo with projects in the monorepo? If you try to answer that question, you will realize we are back to square one.