Movatterモバイル変換


[0]ホーム

URL:


Skip to content

Navigation Menu

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up

Fast Commons Category Inspection

License

NotificationsYou must be signed in to change notification settings

Commonists/fastcci

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

FastCCI Logo

Fast Commons Category Inspection is an in-memory database for fast commons category operations such as

  • Loop detection
  • Deep traversal
  • Category intersection
  • Category subtraction

FastCCI can operate without depth limits on categories.

fastcci_build_db builds the binary database files from an SQL dump of the categorylinks database.

fastcci_server is the database server backend that can be queried through HTTP.

Where is FastCCI used?

An instance of the FastCCI backend is running on Wikimedia Labs athttp://fastcci1.wmflabs.org/. A frontend is available on Wikimedia Commons as a gadget (Click here to install).

Preparing database

The database is generated from a simple parent child pageid table that is generated with a short SQL query. On Wikimedia Tool Labs this query can be launched with the following command.The text output is streamed into thefastcci command that parses it and generates a binary database image, containing of thefastcci.cat index file and thefastcci.tree data file.Both files are saved to the current directory.

mysql --defaults-file=$HOME/replica.my.cnf -h commonswiki.labsdb commonswiki_p -e 'select /* SLOW_OK */ cl_from, page_id, cl_type from categorylinks,page where cl_type!="page" and page_namespace=14 and page_title=cl_to order by page_id;' --quick --batch --silent | ./fastcci_build_db

Query syntax

Start the server with./fastcci_server PORT DATADIR, wherePORT is the tcp port the server will listen, andDATADIR is the path to thefastcci.cat andfastcci.tree files.

The server can be queried through HTTP or WebSockets. The URLs are the same in both cases (except for the protocol part). The request string looks like an ordinary HTTP GET URL.assuming the server was started on port 8080 you can query it using curl like this:

curl 'http://localhost:8080/?c1=9986&c2=26398707&a=path'

For backwards compatibility with browsers that do not properly support cross-domain requests a JavaScript callback mode exists, that wraps the result data in a function call to afastcci_callback function. This mode is activated by adding thet=js query parameter and value.

Query parameters

  • c1 The primary category pageid integer value. This always has to be specified, otherwise the server will return an error 500.
  • c2 The secondary category (or file) pageid
  • d1 The primary search depth (defaults to infinity)
  • d2 The secondary search depth (defaults to infinity)
  • a The query action. Values can be:
    • and Perform the intersection between categoryc1 and categoryc2 (default action)
    • not Fetch files that are in categoryc1 but not in categoryc2
    • list List all files in and below categoryc1
    • fqv List all FPs, QIs, and VIs files (in that order) in and below categoryc1
    • path Find the subcategory path from categoryc1 to file or categoryc2

The server performs some sanity checking on the query parameters to make sure that the pageids supplied are pointing to categories (or if allowed to files).

Response format

The response is delivered in a simple text format with multiple lines. Each line starts with a keyword and may be followed by data. The keywords are:

  • RESULT followed by a| separated list of up to 50 integer triplets of the formpageId,depth,tag. Each triplet stands for one image or category.
  • NOPATH indicates that no path fromc1 toc2 in aa=path request was found.
  • OUTOF followed by an integer that is the number of total items in the calculated result (rather than the number of returned items). This can be either an exact number (fora=list) or an estimate (fora=and anda=not).
  • QUEUED is the immediate acknowledgement that the server has queued the current request.
  • WAITING is sent to the client with one integer value representing the number of requests that are ahead in the queue and will be processed before the current request.
  • WORKING followed by two integers representing the current number of items found inc1 andc2. This response item is sent to the client every 0.2s and shows the current state of the ongoing category traversal.
  • DONE indicates the end of the server transmission.

Command line tools

  • fastcci_tarjan usesTarjan's Algorithm to findstrongly coupled components in the category graph. Those are essentially connected clusters of loops.
  • fastcci_circulartest uses a custom algorithm to find individual category loops. Unlikefastcci_tarjan this also catches self referencing categories. It may however omit loops that share nodes with other loops.
  • fastcci_subcats cat_id outputs the direct subcategories of the category specified bycat_id (this is mostly for debugging).
  • fastcci_pfs_search P F S finds all categories withP parent categories,F number of files, andS subcategories.
  • fastcci_diamond finds all categorydiamonds (i.e. B,C categories with a common parent A and a common subcategory D).

Server setup

Systemd

Copy the service fileassets/fastcci.service to/lib/systemd/system/fastcci.service and symlink it to/etc/systemd/system. Register the service with

systemctl daemon-reloadsystemctl start fastcci.service

Inspect the FastCCi logs withsudo journalctl -e -u fastcci.service

Upstart

Copy the service fileassets/fastcci-server.conf to/etc/init/fastcci-server.conf.


[8]ページ先頭

©2009-2025 Movatter.jp