Discussion:
Sunfish V Lanman98 + a little problem
(too old to reply)
Bob Latham
2019-07-29 09:50:58 UTC
Permalink
I have a library of music in flac format stored on a synology DS214+
NAS. I've written a bit of code, (none multitasking assembler) which
scans the library, lists the contents, opens each and every file and
checks the vorbis tags for certain criteria I prefer.

Just in case it interests anyone..
Running this on Sunfish takes 123219 cs or ~20.5 minutes.
On Lanman98 it takes 80838cs or ~13.5 minutes.
Exactly the same code and library.

But I do have a problem. I think there are still a few, very few I
hope, files in the library that have utf8 characters. My policy is to
change these when I can find them to normal ascii chars but it is
finding them that's my problem.

I have seen two effects of these chars above &80, one is that the RO
filer shows files (text default) but they have no name at all. Worse
though are files that don't show at all in the filer.

Any good ideas (code wise) of how to find these pesky files would be
appreciated.


Thanks

Bob.
--
Bob Latham
Stourbridge, West Midlands
druck
2019-07-29 19:25:43 UTC
Permalink
Post by Bob Latham
I have a library of music in flac format stored on a synology DS214+
NAS. I've written a bit of code, (none multitasking assembler) which
scans the library, lists the contents, opens each and every file and
checks the vorbis tags for certain criteria I prefer.
Just in case it interests anyone..
Running this on Sunfish takes 123219 cs or ~20.5 minutes.
On Lanman98 it takes 80838cs or ~13.5 minutes.
Exactly the same code and library.
Check how fast similar operations work from Linux, to see if there is
any difference in how the NAS handles both network filing systems. I
would expect NFS to be quicker on Linux. If it is it points the finger
at RISC OS's implementation.
Post by Bob Latham
But I do have a problem. I think there are still a few, very few I
hope, files in the library that have utf8 characters. My policy is to
change these when I can find them to normal ascii chars but it is
finding them that's my problem.
I listed the entire RISC OS disc (using a little BASIC dirall program),
sent the output to a file, then used Zap to search of top bit set
characters - and also other characters which aren't mapped symmetrically
by some filing systems, which breaks backups.
Post by Bob Latham
I have seen two effects of these chars above &80, one is that the RO
filer shows files (text default) but they have no name at all. Worse
though are files that don't show at all in the filer.
That's a definite bug.

---druck
Bob Latham
2019-07-30 09:15:13 UTC
Permalink
[Snip]
Post by druck
Post by Bob Latham
Just in case it interests anyone..
Running this on Sunfish takes 123219 cs or ~20.5 minutes.
On Lanman98 it takes 80838cs or ~13.5 minutes.
Exactly the same code and library.
Check how fast similar operations work from Linux, to see if there
is any difference in how the NAS handles both network filing
systems. I would expect NFS to be quicker on Linux. If it is it
points the finger at RISC OS's implementation.
No idea as to the cause of the speed difference, I can only report
what I've seen for anyone interested. I have plenty of Linux devices
which I fiddle with but I don't have a Linux computer.

Something else that is quite odd, the routine I wrote is recursive
and stops when it gets back to the starting level if you see what I
mean. There is a counter running so I know when the last directory
has been scanned. On LM98, the prog stops instantly the last
directory has been printed on screen. On sunfish it is another 20
seconds before the prog stops. What the blazes is it doing in that
time? Rhetorical, I don't expect an answer.


RO 5.25 - LM98 2.06 - sunfish 2.05
Post by druck
Post by Bob Latham
But I do have a problem. I think there are still a few, very few
I hope, files in the library that have utf8 characters. My policy
is to change these when I can find them to normal ascii chars but
it is finding them that's my problem.
I listed the entire RISC OS disc (using a little BASIC dirall
program), sent the output to a file, then used Zap to search of
top bit set characters - and also other characters which aren't
mapped symmetrically by some filing systems, which breaks backups.
Your suggestion made me look at the issue another way and after some
googling I found out how to make W10-DOS list all files in a
directory tree - got it. I've then found out how to create a sub list
that only contains files with top bit set chars using SE.

Your post was a big help, thanks.
Post by druck
Post by Bob Latham
I have seen two effects of these chars above &80, one is that the
RO filer shows files (text default) but they have no name at all.
Worse though are files that don't show at all in the filer.
That's a definite bug.
---druck
Cheers mate,


Bob.
--
Bob Latham
Stourbridge, West Midlands
Loading...