Grep final set of parameters from fit.log gnuplot file

I would like to grep the final set of fit parameters from a gnuplot log file to form columns that look like this.

a_1001 b_1001 x_1001
a_1002 b_1002 x_1002
a_1003 b_1003 x_1003
   .          .         .   
   .          .         .
   .          .         .
a_1250  b_1250 c_1250

At present, the 'fit.log' file looks like this (also see attachment).

*****************************************************************




 Iteration 0
 WSSR        : 13.5219           delta(WSSR)/WSSR   : 0
 delta(WSSR) : 0                 limit for stopping : 1e-05
 lambda	  : 1.1547

initial set of free parameter values

a_1001          = 1
b_1001          = 1
x_1001          = 1

After 8 iterations the fit converged.
final sum of squares of residuals : 0.00106152
rel. change during last iteration : -1.34952e-06

degrees of freedom    (FIT_NDF)                        : 15
rms of residuals      (FIT_STDFIT) = sqrt(WSSR/ndf)    : 0.00841238
variance of residuals (reduced chisquare) = WSSR/ndf   : 7.07681e-05

Final set of parameters            Asymptotic Standard Error
=======================            ==========================

a_1001          = 0.497897         +/- 0.001983     (0.3982%)
b_1001          = 0.00459204       +/- 0.005632     (122.6%)
x_1001          = 1.57077          +/- 0.6132       (39.04%)


correlation matrix of the fit parameters:

               a_1001 b_1001 x_1001 
a_1001          1.000 
b_1001         -0.003  1.000 
x_1001         -0.000  0.002  1.000 


****************************************************************

 Iteration 0
 WSSR        : 13.5246           delta(WSSR)/WSSR   : 0
 delta(WSSR) : 0                 limit for stopping : 1e-05
 lambda	  : 1.1547

initial set of free parameter values

a_1002          = 1
b_1002          = 1
x_1002          = 1

After 8 iterations the fit converged.
final sum of squares of residuals : 0.0019177
rel. change during last iteration : -1.59332e-09

degrees of freedom    (FIT_NDF)                        : 15
rms of residuals      (FIT_STDFIT) = sqrt(WSSR/ndf)    : 0.0113069
variance of residuals (reduced chisquare) = WSSR/ndf   : 0.000127847

Final set of parameters            Asymptotic Standard Error
=======================            ==========================

a_1002          = 0.49692          +/- 0.002665     (0.5363%)
b_1002          = 0.00889775       +/- 0.007585     (85.24%)
x_1002          = 1.57078          +/- 0.4262       (27.13%)


correlation matrix of the fit parameters:

               a_1002 b_1002 x_1002 
a_1002          1.000 
b_1002         -0.006  1.000 
x_1002         -0.000  0.002  1.000 

I only need the final set of parameters as a table and I am wondering if anyone could help using awk or grep to get these parameters out.

I see nothing resembling that output in your input so am confused to which lines you expect grep to match.

For your test input, is this the first line of output you expect?

0.497897 0.00459204 1.57077

Yes, my output should look like this:

0.497897 0.00459204 1.57077
0.49692 0.00889775 1.57078

And so on.. Thanks

Here's one possible way:

$cat temp.sh
sed -n "/^Final/,/^correlation/ p" infile > temp1.x # get segments
grep "^[a-z]_[0-9][0-9][0-9][0-9] " temp1.x > temp2.x # get lines
sed "s/.* = \([^ ][^ ]*\) .*/\1/" temp2.x > temp3.x # get numbers
sed "N; N; s/\n/ /g" temp3.x # join lines

$ ./temp.sh
0.497897 0.00459204 1.57077
0.49692 0.00889775 1.57078

If your sed does not support \n, there is bound to be some other simple way to join lines on last step.

With the sample provided, the following simple awk script seems to work:

awk -F' *=* +' '/[+]\/-/{printf "%s%s",$2,(++i%3)?" ":"\n"}' fit.txt

If you are using a Solaris/SunOS system, use /usr/xpg4/bin/awk or nawk instead of awk .

In your original problem, each entry in fit.txt contained 3 lines that contained the string +/- and the awk script I supplied plucked a value from each of these lines and added a new line after find the 3rd matched line in each entry:

awk -F' *=* +' '/[+]\/-/{printf "%s%s",$2,(++i%3)?" ":"\n"}' fit.txt

Now fit.txt contains entries each of which has 4 lines that contain the string +/- and guess what: the awk script:

awk -F' *=* +' '/[+]\/-/{printf "%s%s",$2,(++i%4)?" ":"\n"}' fit.txt

will do what you want. With you new fit.txt sample file, it produces:

0.498062 -0.00287463 -0.017153 -0.0145765
0.497897 -0.00459203 -0.0162969 -0.0139438
0.49692 -0.00889774 -0.0210222 -0.0182511

See the pattern? This pattern will work as long as the number of variables you're concerned about is a constant for any given input file. The logic would be more complex if the number of fields were to vary between entries.

Thanks Don Cragun. It works like magic!!