Add option to limit Lerna to symlink only packages defined in scope

[ibezkrovnyi]

A - monorepo packages

S - scope packages, list of packages combined together by --scope, --include-filtered-dependents, --include-filtered-dependencies, --since

N - 'not in scope' packages: A - S

Expected Behavior

Lerna symlinks only projects from list S to their dependents from list S.

Current Behavior

Lerna symlinks projects from list A to their dependents from list A

Tested locally

lerna/core/package-graph/index.js

Line 116 in d6e6a42

if (forceLocal || resolved.fetchSpec === depNode.location || depNode.satisfies(resolved)) {

If I change above line into:

const local = forceLocal || resolved.fetchSpec === depNode.location || depNode.satisfies(resolved);
if (local && depName !== 'project-name') {

then node_modules/project-name in all lerna.json packages will not be replaced with symlink.

Solution

We need to filter out some packages from being symlinked

Option 1

To add cli parameter (let's say, --exclude-from-local-dependencies) and pass its argument (list of packages or minimatch patterns) into PackageGraph (for all commands).
But is more like tricky workaround - we will need to manually calculate difference between lerna ls and lerna ls --since --include-filtered-dependents, and pass diff to --exclude-from-local-dependencies

Option 2 (best solution)

To add cli parameter (let's say, --exclude-non-scope-from-local-dependencies, or --scope-local by analogy with force-local) and auto-diff by lerna and pass result to PackageGraph (for all commands). It would be the best solution

Context

We have couple of monorepos with ~200 projects in each, and with current behavior of Lerna we have to manually build almost all our projects just to test couple of them.

Related issues

#1421 - is about --ignore option, and this issue is about scope

Your Environment

Executable	Version
lerna --version	3.4.3
npm --version	6.4.1
node --version	10.13.0

OS
NAME
RedHat
Windows

[olvnikon]

We faced a similar problem. I'm also interested in the solution.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[ibezkrovnyi]

For anyone who will have this issue there is workaround with writing custom script:
#1421 (comment)

It still is not as perfect as native lerna functionality, but at least it solves this issue without touching Lerna code.

[evocateur]

I don't believe altering the initial graph painting is the right place, it should be a method similar to PackageGraph's partitionCycles(), but instead returns a brand new PackageGraph with the unchanged siblings excluded.

[lock]

This thread has been automatically locked because there has not been any recent activity after it was closed. Please open a new issue for related bugs.