User:GPSLeo/stats-tools

Some little Tools to analyze Commons.

Get edit list

#!/usr/bin/env python3

import pywiki # my own toolset for mediawiki API stuff
import json

ua = "DATA-SCRIPT/0.1 (User:GPSLeo)"
url = "https://commons.wikimedia.org/w/api.php"

req = pywiki.setloginBot(ua)

out = []

quest = {
	"action": "query",
	"format": "json",
	"list": "recentchanges",
	"utf8": 1,
	"formatversion": "2",
	"rctype": "edit|new",
	"rcprop": "title|timestamp|ids|tags|patrolled|user",
	"rcshow": "anon",
	"rclimit": "max"
}
#"rctag": "mobile edit"

n = False

while n == False:
    data = req.get(url, params=quest)
    data = data.json()
    print(data)
    for elem in data["query"]["recentchanges"]:
        out.append(elem)
    if "continue" not in data:
        n = True
    else:
        quest["rccontinue"] = data["continue"]["rccontinue"]
    
        
with open("edits.json", "w") as outfile:
    outfile.write(json.dumps(out, indent=4))

Analyze

Script does not run! Only a collection of code lines.

library(jsonlite)
library(plyr)
library(stringr)

setwd("")

data <- fromJSON("edits.json", simplifyDataFrame = T)

mobile <- fromJSON("editsmobile.json", simplifyDataFrame = T)

data$user <- as.factor(data$user)

users <- count(data$user)

#Users with more edits then
users100 <- users[which(users$freq >= 100),]

#Users with edits between
users2to9 <- users[which(users$freq >= 2 & users$freq <= 9),]

#Base stats
nrow(users100) #Number of IPs

sum(users2to9$freq) #Number of edits made by these IPs

# Count number of patrolled and autopatrolled(reverted) edits
usersVect <- users
patrolled = 0
autopatrolled = 0

for(r in 1:nrow(data[data$user %in% usersVect,])){
  if(data[r,"patrolled"] == T){patrolled = patrolled+1}
  if(data[r,"autopatrolled"] == T){autopatrolled = autopatrolled+1}
}

# Smaller variant for all edits
count = 0
for(r in 1:nrow(mobile)){
  if(mobile[r,"patrolled"] == T){count=count+1}
  #if(mobile[r,"autopatrolled"] == T){count=count+1}
}

As for IPs auropatrolled is practically the same as reverted. So we can take the number of auropatrolled edits as the number of reverted edits. The tags are not unsable in R without some cleanup before.
For the patrolled edits we need to reduce the count for edits marked as patrolled by the count of autopatrolled edits.

• Reverted = autopatrolled
• Patrolled = patrolled - autopatrolled
• Checked = patrolled
• All edits = sum of all edits

Results

Results as of 2021-12-15 13:40 and bit later for mobile edits.

	All IPs	IPs with >100 edits	IPs with 10-100 edits	IPs with 2-9 edits	IPs with 1 edit	IP edits mobile
IPs in group	7378	85	449	2299	4545	3803
Edits	54983	29975	13304	7359	4545	7889
% of IP edits	100%	54.5%	24.2%	13.4%	8.3%	14.3%
Unchecked edits	50874	28252	12530	6858	4230	5550
Unchecked edits %	92.5%	94.3%	94.2%	93.2%	93.1%	70.4%
Checked edits (Revert + Patrol)	4109	1723	774	501	315	2339
Checked edits % in group	7.5%	5.7%	5.8%	6.8%	6.9%	29.6%
Patrolled edits	1845	528	211	130	51	673
Patrolled edits % in group	3.4%	1.8%	1.6%	1.8%	1.1%	8.5%
Reverted edits	2264	1195	563	371	264	1666
Reverted edits % in group	4.1%	4%	4.2%	5%	5.8%	21.1%
Reverted edits % of checked in group	55.1%	69.4%	72.7%	74.1%	83.8%	71.2%