browserify for node

Usage no npm install needed!

<script type="module">
  import noderify from '';



browserify for the server side.

Why would you do this?

node modules are great, and make it easy to reuse lots and lots of openly available modules. But node loads these modules synchronously, so if you have an application with many modules, just loading the code may take a significant time.

If you have a spinning disk, this may be very unpleasant, and even if you have a SSD it may still be slow enough to not feel "snappy"

on my computer (ThinkPad X220, with spinning disk)

$ time npm --version

real    0m2.384s
user    0m0.530s
sys     0m0.043s

# okay...

$ time browserify --version

real    0m3.014s
user    0m0.950s
sys     0m0.097s

# too slow...

$ time npmd --version

real    0m7.447s
user    0m1.240s
sys     0m0.157s

# wow too slow

$ time node sbot version # (scuttlebot)

real    0m9.103s
user    0m1.120s
sys     0m0.187s

# WTF?

The reason this is so slow, is because each require blocks, so if you have hundreds you have to wait for each one. lets bundle this into one file, so we only read one file, and then everything is in memory.

$ noderify scuttlebot/bin.js > b.js
$ time node b.js version

real    0m1.038s
user    0m0.553s
sys     0m0.033s

# that is MUCH better! 8.76 times faster!!!


noderify entry.js [flags]

  --filter, -f          ignores a module from the bundle,
                        use this for native add-ons     replaces module "foo" with module "bar"
  --out, -o             specify the output file
  --electron, -e        ignore common electron modules
  --ignoreMissing       ignore modules that were not found
  --verbose             turn on verbose logging
  --help                show this help info

If --out / -o isn't specified the results will be logged to stdout, meaning you can shovel the results into a new file (e.g. noderify entry.js > bundle.js)

Since noderify uses rc it configuration may be set in a local .noderifyrc file in json format.

  "filter": ["module_name1", "module_name2"],
  "replace": {
    "sodium-native": "sodium-javascript"

how it works

noderify creates a javascript bundle with a different arrangement to browserify. the first thing is a prelude function which takes content and structure of the bundle and assembles it together (injecting require and module variables, etc) this function is self-evaluating so running the javascript bundle runs the program.

prelude(content, dependencies, entry)

content is a content addressed mapping from the base64 encoded sha256 hash of each content file to that file (as a module closure) in node.js modules, a given file may be used multiple times, and might even have different names. Since the content is hashed, if different versions of a single module occur, but have different names, they are not duplicated.

content = {<hash(file)>: <file>, ..}

dependencies represents the dependency tree. It is a mapping of filename to the content hash (so it can be looked up in the content object) and then that file's internal dependencies. the file's dependencies map from the dependency expression within that file (it could be relative like require('./lib/util.js') or to an installed module like require('minimist') by pointing to the actual filename that expression maps to, the noderify prelude function does not require an internal model of the file system, it can simply look up the mappings for each file.

dependencies = {
  <filename>: [hash(file), { <relative_require>: <filename>, ...}]

entry is just the file name in the dependency tree to call first.

TODO - dynamic linker

noderify (like browserify) is essentially a static linker for javascript for certain applications, it would be interesting to have a dynamic linker. In particular, for something like depject a dynamic linker could combine multiple dependency trees into a single bundle.