[mythtv] mythfilldatabase, Schedules Direct, json and xmltv

[Peter Bennett]

I have done some testing with the three methods of using the json feed
from Schedules Direct.

I have a lineup that contains 520 channels, of which I actually receive 217.

I am unable to get Robert K's program to work, see
https://github.com/SchedulesDirect/mfdb-json/issues/20

tv_grab_sd_json from XMLTV
The package from the xmltv site works well. I ran it with all channels
selected because the user interface for picking your 217 channels out of
the 520 possible is awful.
It took some 15 minutes to run, at 100% of one CPU for a lot of the time
and used 3.9 GB resident memory. Since my backend has only 4 GB, I
restricted system memory to 4GB and tried again. It took 30 minutes and
used about 2 GB of swap space in the process. The disk was thrashing
terribly. I fear that if I had the backend trying to record at the same
time that the recordings would be damaged or would fail.

tv_grab_na_sd from Gary Buhrmaster
This package works well. It has the advantage that it only downloads as
much as it needs each time. It has the same awful interface for
selecting your channels.
The first download took about 10 minutes. Running mythfilldatabase with
the xmlfile option used 4.98 GB resident memory, but low CPU utilization.
I used a SQL update against the SQLite database to select the channels
that I wanted. On my 217 channels the resident memory used by
mythfilldatabase was 2.436 GB and it took 2 minutes.

With the Schedules Direct DD feed, mythfilldatabase uses around 90 MB.

Using the json feeds gives me three weeks of data, compared to the 12
days I get with the DD feed, also it finds icons for all the channels,
so there are some major improvements.

It would be nice if we could reduce the memory usage of mythfilldatabase
and provide a better way of selecting channels.

Peter