What is the portable way of using NVM's node in Makefiles

[tuaris]

I know how to call NVM from a Makefile, assuming everyone is using NVM. What if there is someone out there that's not using NVM? What's the correct way to use the environment NVM sets up from within a Makefile without assuming NVM exists and without forcing a default version change on the user's NVM environment?

For example, in the shell I can do:

node -v # Outputs 20
nvm use # Gets version from .nvmrc with '18' in it
node -v # Outputs 18
npm install
npm build
....

Consider the following Makefile

build:
    node -v
    npm install
    npm build

I would like it to function as follows:

node -v # Outputs 20
nvm use # Gets version from .nvmrc with '18' in it
node -v # Outputs 18
make build
node -v # Outputs 18
....

[ljharb]

I'm not sure what you're asking.

nvm is a sourced shell function, so for a user to have it available, they have to have installed it on their machine. A Makefile can't set that up for them (short of running the install script)

[tuaris]

I'm not sure what you're asking.

nvm is a sourced shell function, so for a user to have it available, they have to have installed it on their machine. A Makefile can't set that up for them (short of running the install script)

I'd like my example Makefile to work with and without NVM. If the user wants to or has a version of node/npm setup by NVM, the Makefile should respect it.

[ljharb]

Certainly the makefile can which node, and if it's found, just use that - and if not, set up its own nvm and use that?