https://github.com/ceejbot/npm-scored-search
an experiment in using norch/search-index for searching npm modules
https://github.com/ceejbot/npm-scored-search
Last synced: about 2 months ago
JSON representation
an experiment in using norch/search-index for searching npm modules
- Host: GitHub
- URL: https://github.com/ceejbot/npm-scored-search
- Owner: ceejbot
- License: isc
- Created: 2015-05-24T16:13:11.000Z (about 11 years ago)
- Default Branch: master
- Last Pushed: 2015-05-25T15:02:47.000Z (about 11 years ago)
- Last Synced: 2025-02-24T03:56:39.547Z (over 1 year ago)
- Language: JavaScript
- Size: 215 KB
- Stars: 2
- Watchers: 4
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# npm-scored-search AKA scorch
Three components:
1. A pluggable backend for [registry-follower](https://github.com/npm/registry-follower) that stores data in a sharded norch [search-index](https://github.com/fergiemcdowall/search-index) instead of elasticsearch.
2. An adapter for a norch [search-index](https://github.com/fergiemcdowall/search-index) that exposes a simplified npm-centric api for use by...
3. A restify server that exposes a query interface. This first searches by keywords using norch, then, if given an npm user, re-weights package scores using the package.json social graph.
[](https://travis-ci.org/ceejbot/npm-scored-search)
## Requirements
* write access to the norch search index path, with sufficient space
* probably redis for caching personalized scores (eventually)
* probably some kind of tls termination in front
## npm script targets
* `start`: start the service; appropriate for production environments
* `dev`: start the service with pretty-printed console logging & NODE_ENV=dev
* `test`: run tests
* `lint`: run jshint against source files
* `test-travis`: run jshint as well as tests, and enforce some minimum coverage
* `logtail`: tail production logs through a pretty-printer
* `repl`: connect to the repl socket with the default name (socket uses the node name so this *should* be different in production; correct socket name is logged at start)
## backend for registry-follower
TBD
Requires the URI for a running query server as configuration.
(TODO: open-source registry-follower, which is probably of general use once npm-specific backends are stripped out)
## adapter
TBD
Used by the server to wrap the search-index api in a convenient way.
## server
### API exposed
All endpoints expect and respond with json.
* `POST /package`: add a package to the search index; body must be json
* `DELETE /package:id`: remove a package from the search index
* `GET /matches/:prefix`: respond with an array of package names that start with the prefix
* `GET /search/:terms`: uri-encoded search term string is turned into a query; response is an array of packages that match somehow
### Configuration
```
Usage: scorch --name scorch-1 --listen localhost:5555 --redis http://localhost:
6379
Options:
-n, --name node name for metrics & logging [default: "scorch"]
-l, --listen host:port pair to listen on [default: "localhost:5757"]
-r, --redis host:port pair for redis [default: "http://localhost:6379"]
-i, --index path for norch index file [default: "norch"]
-m, --metrics numbat metrics URI; optional
--help Show help [boolean]
```
### monitoring hooks
TODO: open-source restify-monitor, which doesn't do much of anything special or sekrit. (It's now published as `@ceejbot/restify-monitor` as a temp hack.)
* `/_monitor/ping`: responds with 200 'pong' if up
* `/_monitor/status`: responds with 200 and a json object with PID & memory usage information
Also creates a repl. Location of repl socket is logged at start.
### metrics
The query proxy emits the following metrics:
* `scorch.start`: `'pid': process.pid`
* `scorch.latency`: `'value': latency`: on each request served, how long it took
## LICENCE
ISC.