Peculiar results using mfscanf in Scilab 4.1.2

Richard Owlett

unread,

Oct 31, 2009, 11:33:17 AM10/31/09

to

My test code fragment is
datafile = xgetfile('*.log',"G:/2008 downloads/",title = 'Chose
an input file');
fd = mopen(datafile,'r'); // open the file for reading

stride = 1
tmpw = mfscanf(stride, fd, '%s %i %i %*c %i %*c %i %*5c %s ' )
tmpw = mfscanf(stride, fd, '%s %i %i %*c %i %*c %i %*5c %s ' )
tmpw = mfscanf(stride, fd, '%s %i %i %*c %i %*c %i %*5c %s ' )
tmpw = mfscanf(stride, fd, '%s %i %i %*c %i %*c %i %*5c %s ' )
tmpw = mfscanf(stride, fd, '%s %i %i %*c %i %*c %i %*5c %s ' )
tmpw = mfscanf(stride, fd, '%s %i %i %*c %i %*c %i %*5c %s ' )
tmpw = mfscanf(stride, fd, '%s %i %i %*c %i %*c %i %*5c %s ' )
tmpw = mfscanf(stride, fd, '%s %i %i %*c %i %*c %i %*5c %s ' )

In actual application the stride is 1000 and the mfscanf
statement is inside of a loop. As actual data is >300k lines I'm
looking for improvements in speed.[16M lines not unreasonable. If
I get that far I would hope to convince data source to improve
the format ;] My original code had
tmpw = mfscanf(stride, fd, '%s %s %s %s %s ' )
Parsing the result seemed to be my bottleneck so I went looking
for alternatives. The rightmost column of data may have 1 or 2
digits to left of the decimal point.

Comments or suggestions?
TIA

My test data file contents are
Sep 03 13:06:00 freq: 10.87
Sep 04 13:06:02 freq: 10.95
Sep 09 13:06:12 freq: 10.88
Sep 10 13:06:14 freq: 10.92
Sep 11 13:07:13 freq: 10.86
Sep 12 13:07:15 freq: 11.02
Sep 13 13:07:17 freq: 11.05
Sep 14 13:07:19 freq: 11.43
Sep 15 13:07:21 freq: 10.94
Sep 16 13:07:23 freq: 11.17
Sep 27 13:05:51 freq: 10.92
Jan 18 13:05:53 freq: 10.90
[above is cut and paste from my editor]

When I paste my test code fragment to the Scilab window I get

Startup execution:
loading initial environment

-->datafile = xgetfile('*.log',"G:/2008 downloads/",title =
'Chose an input file');

-->fd = mopen(datafile,'r'); // open the file for reading

-->

-->stride = 1
stride =

1.