Minor bug reading null/empty series or episode #s from XMLTV file EPG source?

[sjabaker]

I believe I've spotted a minor bug in the xmltv file EPG import process: NPVR *appears* to be assuming Series 01 if there's an episode number but no series number in the xmltv_ns episode_num field, and assuming Episode 01 if series # but no episode # present. For example:
<episode-num system="xmltv_ns">75 .  . 0/1</episode-num> input is showing up in the EPG as 'S76E01'
<episode-num system="xmltv_ns"> . 7/8 . 0/1</episode-num> input is showing up as 'S01E08'

I'm pretty sure it's not actually an assumption being made (I know sub's noted in other threads that he doesn't want to do that as it's so frequently wrong) - rather I suspect it's the result of a string-to-number conversion that's producing '0' for the null/empty input, which given the zero-based numbering of the xmltv format is then showing in the EPG as S01 or E01.

(No such issue occurs with the direct DVB/ATSC EPG, when compared to importing from an xmltv file produced from the same original DVB source.)

Regards,
Steve

[mvallevand]

I understand what you are saying about the discrepancy with assumptions, but is the assumption actually wrong in the cases you found?

Martin

[sjabaker]

Yes.

The vast majority are where just an episode # has been derived from data in the description such as "Ep 5" or "3/12." While some of these do/will indeed happen be series 1, lots don't/won't. I don't recognise most of the show titles, certainly not well enough to know whether it's season 1 being broadcast or not - but there are some obvious ones such as:
I'm pretty sure all the "Masterchef" episodes would be the latest season and not s1 (from 2005 according to tvdb) being repeated, and similarly for "Live at the Apollo".
"Have I got New for You" is in the middle of series 60-something so the new episodes of that aren't s1, and it's extremely unlikely that any of the repeats on other channels are either.
"Killing Eve" is the last series just about to finish (s4 or 5 I think?), and the "Doctor Who" episode I spotted is from the latest/current set (which is way beyond series 1 regardless whether you're starting from 1960-whatever or only the 2005 reboot).

It's tricky to get any kind of accurate proportions for a more analytical view, since all I can do is lookup the show names individually on something like the tvdb and see if it's had only one series (in which case presumably what's being aired is indeed series 1!) or more than one (in which case by far the most likely scenario is that it's not series 1 being shown).
Based on a small, semi-random sample (ordered by show title so it's not all from the same channel, for example) of the first 20 shows for which I could get what looked like a solid/correct match on TVDB (and assuming their listing of the # of series is correct) I found it to be approximately 50/50 between them (11 multi-season shows vs 9 single-season).

The very very few with only a series number (plus a few of those with only an episode #), are actually ones where the number is wrong in and of itself (due to an over-zealous match in the program producing the source xmltv file) - so technically any assumption is by definition wrong. However there's certainly nothing in the data to even suggest "1" as the right guess for the other number.

[sub]

I’m away for a few days so can’t check it right now. If you bump the thread tuesday, I’ll take a look at code.

[sjabaker]

No worries! - and thanks.
Will do.

[sjabaker]

Just bumping the thread, as requested...

[sub]

I believe I've spotted a minor bug in the xmltv file EPG import process: NPVR *appears* to be assuming Series 01 if there's an episode number but no series number in the xmltv_ns episode_num field, and assuming Episode 01 if series # but no episode # present. For example:
<episode-num system="xmltv_ns">75 .  . 0/1</episode-num> input is showing up in the EPG as 'S76E01'
<episode-num system="xmltv_ns"> . 7/8 . 0/1</episode-num> input is showing up as 'S01E08'

I'm pretty sure it's not actually an assumption being made (I know sub's noted in other threads that he doesn't want to do that as it's so frequently wrong) - rather I suspect it's the result of a string-to-number conversion that's producing '0' for the null/empty input, which given the zero-based numbering of the xmltv format is then showing in the EPG as S01 or E01.

Have you got a xmltv file I could try here?

I think I see where it might be happening.

[sjabaker]

Sure - an up to date xmltv file is in a shared folder at https://1drv.ms/f/s!AtsVBvATpsLhgYmeOlRbVRTn5Kl5H_c
The file is the only .xml file there: "490 MHz 270422 231650 EPGC FP32v3 xmltv.xml"

Some example programmes to see the effect are...
Episode only, "assuming" S01: "Come Dine With Me", "MasterChef", "The Bidding Room"
Series only, "assuming E01: the "Grand Tourer/Ferrari Roma" episode of "Secrets of the Supercars"
(Note that the series-only ones are in fact "incorrect" matches in the created xmltv data, but they still demonstrate the effect if no episode # is present in the xmltv_ns numbering.)

If you want to find others, you can either search the .xml file or just use the .csv file of the same name in the folder, and filter the "From Tag" column for "begins with E" or "begins with S and doesn't include E". There's a total of 1355 programmes with only Episode data, and 5 with only a series #.

Regards,
Steve

[sjabaker]

Not sure if you managed to take a look at that xmltv file - in case not, a new up-to-date file is available here

As an example, you can clearly see that both S01 and E01 are being assumed since the following XML entry - which has only 'part X of Y' data but nothing (i.e. null/blank) for both series and episode - shows up in the NextPVR EPG after ingestion as "S01E01"...
  <programme start="20220607000000 +0100" stop="20220607010500 +0100" channel="41248">

    <title>Air Aces</title>
    <sub-title>Red Tails</sub-title>
    <desc>How pressure from activist groups pushed President Franklin D. Roosevelt to initiate an African-American fighter pilot training programme during World War II. (Part 4 of 6)</desc>
    <category>Arts and Culture</category>
    <episode-num system="xmltv_ns"> .  . 3/6</episode-num>
  </programme>

[sub]

Sorry, I forgot to say that this has been fixed for the next build.