You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
84 lines
2.9 KiB
84 lines
2.9 KiB
#!/usr/bin/env python
|
|
from __future__ import print_function
|
|
from argparse import ArgumentParser
|
|
import sys, json, re
|
|
from datetime import datetime
|
|
from urllib import urlencode
|
|
from urllib2 import urlopen, HTTPError, URLError
|
|
from csv import writer
|
|
from math import ceil, floor
|
|
|
|
"""
|
|
Dumps a CSV of all pads with columns
|
|
padid, groupid, revisions, lastedited, author_ids
|
|
|
|
padids have their group name trimmed
|
|
groupid is without (g. $)
|
|
revisions is an integral number of edits
|
|
lastedited is ISO8601 formatted
|
|
author_ids is a space delimited list of internal author IDs
|
|
"""
|
|
|
|
groupnamepat = re.compile(r"^g\.(\w+)\$")
|
|
|
|
out = writer(sys.stdout)
|
|
|
|
def jsonload (url):
|
|
f = urlopen(url)
|
|
data = f.read()
|
|
f.close()
|
|
return json.loads(data)
|
|
|
|
def main (args):
|
|
p = ArgumentParser("outputs a CSV of information all all pads")
|
|
p.add_argument("--padinfo", default="padinfo.json", help="padinfo, default: padinfo.json")
|
|
p.add_argument("--zerorevs", default=False, action="store_true", help="include pads with zero revisions, default: False")
|
|
args = p.parse_args(args)
|
|
|
|
with open(args.padinfo) as f:
|
|
info = json.load(f)
|
|
apiurl = "{0[protocol]}://{0[hostname]}:{0[port]}{0[apiurl]}{0[apiversion]}/".format(info)
|
|
data = {}
|
|
data['apikey'] = info['apikey']
|
|
requesturl = apiurl+'listAllPads?'+urlencode(data)
|
|
|
|
padids = jsonload(requesturl)['data']['padIDs']
|
|
padids.sort()
|
|
numpads = len(padids)
|
|
maxmsglen = 0
|
|
count = 0
|
|
out.writerow(("padid", "groupid", "lastedited", "revisions", "author_ids"))
|
|
for i, padid in enumerate(padids):
|
|
p = (float(i) / numpads)
|
|
percentage = int(floor(p*100))
|
|
bars = int(ceil(p*20))
|
|
bar = ("*"*bars) + ("-"*(20-bars))
|
|
msg = u"\r{0} {1}/{2} {3}... ".format(bar, (i+1), numpads, padid)
|
|
if len(msg) > maxmsglen:
|
|
maxmsglen = len(msg)
|
|
sys.stderr.write("\r{0}".format(" "*maxmsglen))
|
|
sys.stderr.write(msg.encode("utf-8"))
|
|
sys.stderr.flush()
|
|
m = groupnamepat.match(padid)
|
|
if m:
|
|
groupname = m.group(1)
|
|
padidnogroup = padid[m.end():]
|
|
else:
|
|
groupname = u""
|
|
padidnogroup = padid
|
|
|
|
data['padID'] = padid.encode("utf-8")
|
|
revisions = jsonload(apiurl+'getRevisionsCount?'+urlencode(data))['data']['revisions']
|
|
if (revisions == 0) and not args.zerorevs:
|
|
continue
|
|
|
|
|
|
lastedited_raw = jsonload(apiurl+'getLastEdited?'+urlencode(data))['data']['lastEdited']
|
|
lastedited_iso = datetime.fromtimestamp(int(lastedited_raw)/1000).isoformat()
|
|
author_ids = jsonload(apiurl+'listAuthorsOfPad?'+urlencode(data))['data']['authorIDs']
|
|
author_ids = u" ".join(author_ids).encode("utf-8")
|
|
out.writerow((padidnogroup.encode("utf-8"), groupname.encode("utf-8"), revisions, lastedited_iso, author_ids))
|
|
count += 1
|
|
|
|
print("\nWrote {0} rows...".format(count), file=sys.stderr)
|
|
|
|
|