Menu
▾
▴
#44 Duplicate Pub Tags being generated
The "random characters" added to Tags aren't that random and for titles like "Doctor Who and the Day of the Daleks" (1979) there is an extremely finite set of possibilities, a lot of which are along paths that are exhausted. In this case the default tag is DCTRWHNDTH1979 but that's taken. The software then cycles through 20 possible consonants to use as the last letter i.e. DCTRWHNDTB1979 through DCTRWHNDTZ1979 (missing out DCTRWHNDTY1979). These are all taken. Then it gives up using the letters and uses the 'try' number, for DCTRWHNDT201979. That's taken. For some reason it doesn't reset the try number so it doesn't start at 'B' again: and it doesn't recognise that it's added two characters rather than one, so it keeps the first digit and tries DCTRWHNDT2211979. That's taken. Next is DCTRWHNDT22221979, DCTRWHNDT222231979, DCTRWHNDT2222241979, etc. It rapidly runs out at DCTRWHNDT22222222223333333371979. The oddity is that the live ISFDB can then go on to DCTRWHNDT22222222223333333338197 so there's seems to be a truncation applied there that isn't in the code from Sourceforge. But the truncation just leads to duplicate tags and "lost" pubs.
Discussion
Anonymous
Submitted mod/ca_new.py 1.7 which should alleviate the problem.
Implemented in r2009-04.
Re-opened the bug as per Swfritter's report:
After entering 537 issue of "Strange Horizons" I ended up with 101 system-generated duplicate tags = "STRNGHRZN22222222223333333338200" and 13 = "STRNGHRZN22222222223333333338201". Plus there were some that I cleaned up earlier. Slightly sub-optimal. Interesting to note that all of the duplicate tags appeared in issues with a month greater than 8. I will leave the data as is for a short time if anyone wants to analyze it. Then it is on to cleanup time.--swfritter 13:12, 28 February 2011 (UTC)
It's Bug 2795822 again, but in a different module - pa_new.py at least. BLongley 16:25, 28 February 2011 (UTC)
pa_new.py v1.12 created, which hopefully will fix this problem for good. BLongley 16:51, 28 February 2011 (UTC)
Hopefully the impact of this issue will go away as I implement bug 3153982 (which I am interpreting as removing all code use of pub tags). With the latest patch accepted, the code already does not link to pubs using pub tags now. It is used for some other things, some of which are sticky to change, but I am working on it.
Fixed in mod/pa_new.py 1.12, implemented in r2011-17.