Thanks. Dependency-tree IS in the memory, but that costs very little. The key is, every node transforms files(called Stream in panto.js) holds a copy of file content, that may takes a lot of memory. My solution is, if the size of content exceed 1Mb, save it to tmp file. I'm thinking about convert strings to buffer in the near future. And the stream idea is great, like gulp does, but I need to merge multiple processes into one. I'm considering how to implement that.
j
stuff ;)
just out of curiosity you keep the dependency-tree in memory I guess ? how does it affect the memory and the cpu ? I'm just curious and this is about metrics not to offend.
I like the idea of 1 source to reduce the IO.... did you solve it via generator streams to reduce memory and only pipe the actual stream buffer through different processes, not sure if JS really supports stream generators atm. but you could build a class for that I guess.