Where does node.js put its files?

Question

I have recently started playing with node.js, but I got lost in a big mess of different versions of node, npm, nvm and other packages. I don't know what is installed globally and what is installed locally (and if locally, how do the packages know which versions of node they can use?).

I'd like to have some summary of what different installation options do. In specific:

• Where is node installed when I use nvm, apt-get, make install or when using other ways?
• Is it a good idea to install node locally?
• Why does nvm change my ~/.profile instead of installing itself in some system-recognizable bin folder?
• I saw that nvm can install different versions of node alongside each other - why would I want to do this? I can install them locally instead, right?
• Where does npm install packages? I saw that it checks packages aganist version of Node, what happens to these packages when node is upgraded?
• In what cases it is better to use global or local installation? Where should I put my packages then (and where they put by default?)
• What's the difference between npm, nvm and nave?

EDIT: There is a lot of ways to install node here, this makes me even more confused...

Answer 1

Where is node installed when I use nvm, apt-get, make install or when using other ways?

apt-get installs all the software, not only node, on the file system following the Ubuntu convention where to store binaries, man files, shared files, logs, etc. However, using apt-get you'll have only the certain version of node which is determined by the distribution release cycle. If there are updates available they will be installed with apt-get update; apt-get upgrade However, the newest version of some app won't be available until it makes its way into the distribution. For example node v0.x.y might not be available until Ubuntu 13.10 the only way to get will be to install it manually. The good side of apt-get or other system package manager is that it manages updates and package removal for you. It stores all the data about the software package in it's own database. You can always delete the node with apt-get remove node and that's it.

make install install the package manually, but it is considered harmful. Never use the make install mainly because you won't be able to delete the package easily, you'll have to read the Makefile and manually delete all the files installed by it. In a situation where you want to use make install there is always checkinstall available. It's a software which creates a native package and registers it with the system. When you decide to delete the package you could do this with one command instead of many. wiki link; Ubuntu guide on checkinstall

Now nvm script is a node version manager. It is very helpful and easy to use. It allows you to have multiple versions of node to be installed and used in parallel on your machine. It doesn't compile the node from source like make install so it is very fast. it doesn't depend on your distribution release cycle so you have access to all the node versions available at the moment. nvm downloads precompiled binaries and is perfect for general use. It stores it's node files in it's own folder locally so in case you want to compare something between the different node versions it's easy to do.

Is it a good idea to install node locally?

If by locally you mean using nvm then it's very good for development, and testing. Not sure about production performance implications and benefits between having it's installed from source or using the nvm precompiled binaries. I use nvm for development and installed from source in production. However if someone could explain this issue any further I'll be glad to learn more.

Why does nvm change my ~/.profile instead of installing itself in some system-recognizable bin folder?

Because nvm isn't an executable. It is a set of bash functions which are sourced by shell and could be used separately. You can invoke nvm_ls and nvm_ls_remote and others without the main script after is is sourced into your shell. What the main script does it parses the command line arguments and pretty prints the output in case of for example `nvm_ls_remote'.

in the ~/.profile the following line is added

[[ -s /home/USERNAME/.nvm/nvm.sh ]] && . /home/USERANME/.nvm/nvm.sh # This loads NVM 

loads all the functions into your shell

I saw that nvm can install different versions of node alongside each other - why would I want to do this? I can install them locally instead, right?

You can install them locally using make install or checkinstall but you will have to make aliases for them like node_0.8.1, node_0.8.2, node_0.10.1 , etc. AND you'll have to manage new aliases, installing all the packages, removing them in case you don't need them YOURSELF. These are a tedious and boring tasks which could be error prone sometimes. nvm does all of these tasks for you for free.

You want to do this to test your app under the different versions of node. For example you are good and tested under the v0.8 but you want to use the new features of the v0.10.3 how do you do that ? You have to download the source code, compile, make an alias and run your app. you could do this with just nvm install 0.10.3 and run your app. Sometimes you have to support more than one version of node. For example some hosted environments are not keeping in touch with the latest release and only have v0.6 Your clients which use your server app might encounter a bug specific to this version. When you fix the bug you have to reproduce it first. Using nvm installation of the v0.6 is one line and half a minute. And you can check all the versions you want this way easily. Test your code under different versions and make sure you are good to go.

Where does npm install packages? I saw that it checks packages aganist version of Node, what happens to these packages when node is upgraded?

If you are using nvm the packages which are installed globally with -g option are tied to the relevant node version. When you switch between versions with nvm use 0.x you have to either install the packages again or use nvm copy-packages <version> to use the packages from in the current version. If the packages are installed locally then it depends. package.json should contain info on the dependencies of the app. If it says node: '0.8' and you just updated to 0.9 you might encounter troubles. For example the behavior of process.nextTick was changed in the latest releases compared to 0.6. So be careful.

In what cases it is better to use global or local installation? Where should I put my packages then (and where they put by default?)

It depends. For development nvm is superior in my opinion. For me it is convenient and simple. For production there are might be some performance implications when using the precompiled binary files not optimized for your system. It would be better to ask this as a separate question so the people with the relevant experience could answer.

What's the difference between npm, nvm and nave?

npm is a node package manager -> link It contains userland packages developed by other people. These packages are not part of the node core. npm is used for publishing your code and dependency management. If your app requires other app developed by other people it is convenient to publish it via npm.

nvm is a node version manager it does a completely separate thing. It gives you an ability to very easily switch between node versions on the same machine and manages all he changes in your $PATH environment variable.

Consider nvm as update manager for the Operation System and npm as a manager of the applications for this system. Well, this comparison isn't precise but just came upon my mind

nave is basically the same as nvm but it is an executable whereas nvm is a script which is sourced into the shell. Each system has it's own benefits. You could make a separate question regarding it's use cases and differences.

My answer isn't 100% complete and contains a lot of subjective personal opinions. However, I hope I'll at least make some points more clear so you might proceed with other more specific questions. Btw, this question list of yours could be asked as separate questions. I believe stackoverflow gives best results when specific questions are asked separately and more people with relevant experience could contribute.