当子模块构建失败时，NPM为什么会成功？

Question

Often while using npm I've come across errors that appear to mean nothing - Visual Studio projects failing to build, build tools (eg: python.exe / CL.exe) not being available on the command line etc.

Some examples of packages I've seen fail to build many times:

• kerberos
• node-gyp
• bcrypt

These throw big error messages with stack traces etc to the console during npm install , clearly having failed completely; however, NPM carries along happy as Larry and 9 times out of 10 my Javascript application and all its dependencies work fine.

1. Does npm install re-build every single dependency recursively, using whatever compilers are available on the local machine?

2. If so, and considering the huge number of dependencies even simple packages can have, how am I able to do ANYTHING without a full suite of programming languages and compilers installed?

3. Why is it that these dependencies failing doesn't necessarily mean my final project will be unusable?

4. If a dependency failing to build is "ok", why bother having the dependency at all?

I haven't been able to find clear answers on any of this, due to the overwhelming number of resources found when searching for terms like "npm build fail".

Answer 1

npm will succeed if those dependencies are actually marked as optional. The ws module is an example of this where they have optional dependencies on two compilable addons. If they fail to build, then ws just uses pure js fallback implementations.

The reason that addons are sometimes added as optional dependencies is that the they (more often than not) perform faster than the pure js alternatives, even for something as "simple" as UTF-8 validation or XOR'ing the contents of a Buffer.