Gzemnid

Data structures

All data files are stored inside the ./pool/ dir by default.

byField.info.json

A huge JSON with basic metadata for all packages on npm. It is an array of entries, the JSON is formatted to be one entry per line.

Includes package name, basic user info, latest version number, npm version used to publish it, and a link to the latest version archive.

Example entry:

{"id":"gzemnid-0.0.1","name":"gzemnid","version":"0.0.1","url":"https://github.com/ChALkeR/Gzemnid/issues","user":{"name":"chalker","email":"[e-mail]"},"npm":"3.9.2","node":"6.2.0","tar":"http://registry.npmjs.org/gzemnid/-/gzemnid-0.0.1.tgz"},

Typical size — under 100 MiB, typical build time — 1-5 minutes, depending on the download speed. It requires downloading about 500 MiB (not stored) in order to build this file.

Created via gzemnid fetch.

This file is required by most of the other commands, so updating any data should begin with re-downloading this file.

stats.json

Contains downloads/month stats for all packages. It is an object from package name to downloads count, the JSON is formatted to be one entry per line.

Example entry:

 "bluebird": 9739667,

Typical size — under 10 MiB.

Created via gzemnid stats.

Requires: byField.info.json (created automatically if not present).

Directory: meta/

Contains files named ${package_name}-${package_version}.json with more detailed package info obtained from the registry, including dependencies for each published version of the package.

One file per ${package_name} is stored, only for the latest version. The JSON files are not formatted.

Example content: see https://registry.npmjs.org/qmlweb.

Notice: these files are updated only on latest version releases, so they might become stale when it comes to beta releases.

Typical size is about 4 GiB, typical bootstrap time is quite long (could even take a day or two), but further updates are quite fast.

Created via gzemnid meta.

Requires: byField.info.json (created automatically if not present).

Directory: current/

Contains latest versions of all packages, one file per package.

Files are named ${package_name}-${package_version}.tgz.

Typical size is currently 95 GiB (and growing), typical bootstrap time depends on your internet connection, further updates are quite fast.

Created via gzemnid packages.

Requires: byField.info.json (created automatically if not present).

Commands

The main script is invoked as gzemnid command [subcommand] (or ./gzemnid.js command [subcommand]), where [subcommand] is optional.

Here is the list of the current commands:

• gzemnid fetch — builds byField.info.json.
• gzemnid stats — runs subcommand rebuild.
  • gzemnid stats rebuild — rebuilds stats.json, downloading stats for all packages present in byField.info.json.
  • gzemnid stats update — updates stats.json for only newly added packages, keeping the numbers for already present packages.
• gzemnid meta — builds meta/ directory, downloading meta info for all packages present in byField.info.json. Outdated files that were present in the meta/ directory are moved to meta.old/.
• gzemnid depsdb — runs subcommands plain, resolved, nested,
  • gzemnid depsdb plain —
  • gzemnid depsdb resolved —
  • gzemnid depsdb nested —
  • gzemnid depsdb stats —
• gzemnid packages — builds current/ directory, downloading latest versions for all packages present in byField.info.json. Outdated files that were present in the current/ directory are moved to outdated/.
• gzemnid extract — runs subcommands partials, totals,
  • gzemnid extract partials —
  • gzemnid extract totals —
• gzemnid code search {regex} — performs a code search over a specified regular expression using the pre-built dataset.
• gzemnid ast execute {file.js} — performs an AST search using the pre-built dataset. Example script — in examples/ast_status.js, execute with gzemnid ast execute ./examples/ast_status.js,
• gzemnid server — starts the web server providing the search API endpoints.

Server

TODO: document server.

Started via gzemnid server.

Deception

Note: think twice before relying on the data obtained from Gzemnid or using it to decide on something.

Code search has both false negatives and false positives — some files are ignored, some files are unused, and some lines could be in a middle of a comment block. Also, your regexps are never ideal.

AST tree also ignores a list of excluded files and directories and minified code and includes unused code and files if those are present in the package for some reason.

Downloads/month are not equal to popularity, and you can't see which version is being used.

Code and AST search, among other things, takes only latest released package versions into an account. That could be significantly different from master, beta branches, also older versions could be much more popular that latest.

All datasets get out of date the moment you build them.

Scoped packages are ignored completely.

Gzemnid deceives you, keep that in mind. But it's still better than nothing.