Status: In development - see the issues
Check the issues to see what needs doing - it'd be great if you'd like to contribute, just pick an issue. The older issues probably need to be done first, just because they were created so things are developed in a sensible order.
Notepad++ Plugin Manager Admin System
This is the web administration interface for Notepad++ Plugin Manager
Plugin authors and maintainers will use this interface to edit their listings within the plugin manager.
The system that has served us well for several years is showing its age, and a number of things have come to light that we weren't aware of when we first built the original system. So, time for a redesign.
What does it do?
Each plugin has a set of metadata (name, author, web site, dependencies etc), and a set of "steps" required to install the plugin. There are currently 4 types of step available - "Download", "Copy", "Run" and "Delete".
"Download" downloads the given URL to a sandbox area, and if it's a
file, will extract all the files under that sandbox.
"Copy" copies a file from the sandbox to one of 3 areas - the plugins directory, the Notepad++ directory, or the plugin configuration directory.
"Run" runs a file, and is rarely used, but can be used install more complex plugins.
"Delete" removes a file from one of the 3 areas that copy can copy to.
A very usual plugin installation consists of a download step to download a ZIP file that contains the DLL of the plugin, and a copy step to copy the DLL to the plugins directory. Some plugins have additional copy steps to copy supporting DLLs etc into other directories.
A plugin author or maintainer can login to this administration system and edit their plugin's information (we trust our users, so currently every user can edit every plugin).
Once the plugin details are saved, a new XML file can be generated, and installed in the host where the Plugin Manager can download it from. This XML is called the development list. In the present setup, this file is automatically available within a few minutes of saving the plugin.
Then, every 2 weeks, the current set of plugin metadata is published as the "live" file. This is obviously problematic, as it means that edits made even minutes before the publish will be published as "live" to (literally) millions of users.
This new implementation stores the data in CouchDB, and each save saves a new instance of the
plugin, complete with all steps etc (previously this was split up to rows in a MySQL database).
Each plugin is actually saved at least twice in Couch, once under the plugin ID, and once under a
unique ID, indexed by the plugin ID. This means all history can be easily retrieved (this won't be
part of the MVP, but is possible in the future). The "development" XML file can then be generated
at any point from the current versions of the plugin definitions. When a plugin is ready to be
published, the plugin is saved again under a further key of the plugin ID with
In a similar way to the edits, the plugin is also saved under a unique key as a
This means a published plugin will have at least 4 "documents" in couchDB - the current edit, the current published, and at least one edit history and one publish history.
This is done so we can always extract a list of published plugins and a list of the "current edits" of plugins, and also always rollback to a previous version.
A copy step can optionally validate that a file is a known file. This is done generally for any executable code, DLLs for example. The MD5 hash of the file being copied is validated against a known list of "good" hashes on the server. This is done at the point of copying the file.
The list of valid hashes is also maintained by this administration system. The old system simply kept a list of valid hashes, which meant it grew forever, and old hashes could never be removed, as it could not be easily known if the hash was still in use.
This system will keep the valid hashes with the saved plugin (i.e. in the plugin document stored in CouchDB). When a file copy step is added in the UI, we can keep track of which files are validated, and update the list of valid hashes for this plugin,in this version.
The project is setup to use Docker to create a local CouchDB instance. This works on Mac and Linux, but has not been tried on Windows. If you would like to contribute and have problems with this step, please raise an issue.
The project uses yarn for package management. Find out from the website how best to install it on your platform if you've not got it installed yet.
Starting the development servers
In order to start the server, simply clone this project, do a
yarn install, and then run
You need to have docker installed and running for this to work.
$ git clone firstname.lastname@example.org:bruderstein/npppm3 $ cd npppm3 $ yarn install ... $ npm start ...
This creates the CouchDB server, installs the necessary design documents (or updates them if they're already there)
and creates an admin user for CouchDB (password is
secret), and an admin user for the website (also
secret). It then starts the node.js server running on port 5003.
In a second terminal,
cd to the
client directory, do a
yarn install there (they have separate
package.json files) and then run
$ cd npppm3/client $ yarn install ... $ npm start
This starts the webpack-dev-server, which listens on port 5001, and proxies requests for
/api through to
the node server listening on port 5003.
Open your browser at http://localhost:5001/ and behold the beauty!
secret, and then you will be presented with an empty list of plugins.
(This is at the time of writing just a white page with a header).
If you navigate to http://localhost:5001/plugins/new you will be able to create a new plugin.
Only the name, author and description fields are mandatory. Once you've saved the plugin, you can
go back to
/plugins and see your new plugin in the list.