[mythtv] myth.rebuilddatabase.pl with new filenames

[Carl Reynolds]

Michael T. Dean wrote:

> <snip... >
> _Proposal_:
>
> <snip... >
>
> IMHO, we should create a script that either a) accepts info as 
> command-line arguments and/or b) prompts the user for info.  (Whether 
> it accepts command-line args or not, it needs the ability to prompt 
> the user for info--otherwise, we're forcing the user to deal with 
> things like maximum command-line length and shell-based 
> quoting/escaping issues.)  It would be possible to allow the user to 
> easily provide information by asking for channel id--even listing the 
> available channel ID's, channel numbers, and channel names and 
> allowing the user to select an option or type their own).  Then, the 
> script should ask for the start date/start time (allowing the user to 
> provide a complete start time or simply the date).  Then, if the user 
> provided a chanid/date that exists within the data in oldrecorded, 
> display a list of shows from that date for the user to select or allow 
> the user to specify a different time.  If the user selected a show, 
> verify the info and then insert it into the db (and update the 
> recstatus in oldrecorded).  If not, then continue prompting for the 
> additional required information.
>
> Comments?  Suggestions?
>
Sorry for taking so long to reply to you. My hard drive filled up on 
Wednesday and I have been trying to manage the data on it since then:

It seems to me that using File::Find::Rule instead of glob to search for 
files would solve most of the problems you have mentioned. We could hard 
code a list of extensions to look for, or the user could provide a list 
of extensions in an ASCII file. File::Find::Rule can be told to filter 
out directory names and files below a certain size and will look in any 
sub-directories. I haven't thought of files existing on other machines. 
Don't know how to solve that one yet.

The user would still need to be prompted for the channel-id, and start-, 
end-date information.

When dealing with massive amounts of files, I don't want to have to type 
the name of each I want to process at a command line prompt. Cutting the 
info down to a [Y/n] prompt or no prompt at all is ideal, but since the 
new naming scheme leaves certain critical information out of the file 
name, I don't see how to eliminate the prompt all together.

It might be possible to derive the start and end dates by looking at the 
creation and last modified dates for the file, but those are just 
guesses since these dates would not be accurate if the file was 
transcoded (for instance). But, even if we can derive the start and end 
dates from the file times, we would still have to prompt the user for 
the channel-id.

It might also be possible to allow an option for the user to pass the 
name of an ASCII file to the script. The file would contain a list of 
the file names and associates channel-ids and start-, end-dates. 
Creating the list of files would be fairly straight forward for the 
average user using 'ls' and an editor, but then entering the channel-ids 
and dates could get really cumbersome because the script would have to 
allow for some variance in personal style.

I'm not sure what you intend to accomplish by re-writing the entire 
script, the basis of what needs to be done is contained in the current 
script and the user/machine interfaces need to be tweaked to make the 
script perform properly and easier to use. Could you tell me why you 
think a total re-write is necessary? Also comment on my proposals above 
and let's see if we can't work out a more usable interface.



Carl.