opus-i2s compatability

[ndeutscher]

Following on from issue #1, and once the opus-i2s.in files can be found, I have struck a series of problems getting the software to work with our files. These work just fine with opus-i2s itself, so it would be good to increase the flexibility here to handle that, as others might also strike similar issues.

I don't know if we want these listed as separate issues, as they are all related, but so far:

• nonlin5 does not support including the path in the individual spectrum listing.
• nonlin5 does not support run numbers with lengths greater than 4 digits (we use 6, and opus-i2s supports this). This means that the outputs of pre-nonlin5 create files such as 'nonlin5_20140101_****.inp' (with the asterisks and inverted commas).

I have yet to code solutions to these issues (and there are possible work arounds), but I will try to do so.

[ndeutscher]

Some progress here, based on recycling what works for i2s.f and associated programs.

The 4 digit issue was easily solved, by increasing the length of run_str and input_file_name in pre-nonlin5.f90, and correspondingly changing the line where the run from the .in file is written to run_str:

•                   WRITE(run_str, '(I4.4)') run
  

•                   WRITE(run_str, '(I6.6)') run
  

•   CHARACTER(LEN = 4) :: run_str
  

•   CHARACTER(LEN = 25) :: input_file_name
  

•   CHARACTER(LEN = 6) :: run_str
  

•   CHARACTER(LEN = 29) :: input_file_name
  

I also changed the definition of run to INTEGER(8).

(updated code to follow, after a separate post to detail fixes for the path in each entry of opus-i2s.in)

[ndeutscher]

To handle the option '0' for input parameter 1 in opus-i2s.in, I borrowed the following changes from the GGG2020 version of i2s.f:

@@ -29,7 +29,7 @@ PROGRAM pre_nonlin5

    CHARACTER(LEN = 14) :: run_file
    INTEGER(KIND = 4), PARAMETER :: iou3 = 13

•   INTEGER(4) :: rs
  

•   INTEGER(4) :: rs,lb,lf,fnbc,fbc^M
  
    CHARACTER(LEN = 21) :: script_file
    INTEGER(KIND = 4), PARAMETER :: iou4 = 14
  

@@ -45,11 +45,12 @@ PROGRAM pre_nonlin5
INTEGER(4) :: l
INTEGER(4) :: i2s_par_no

•   CHARACTER(LEN = 21) :: spectrum_name
  

•   CHARACTER(LEN=100) :: spectrum_name^M
  

•   CHARACTER(LEN=:), allocatable :: inpstr^M
    INTEGER(4) :: year
    INTEGER(4) :: mon
    INTEGER(4) :: day
  

•   INTEGER(4) :: run
  

•   INTEGER(8) :: run
    REAL(KIND = 8) :: lat
    REAL(KIND = 8) :: lon
    REAL(KIND = 8) :: alt
  

(below has some debugging write statements removed)
@@ -187,10 +188,14 @@ PROGRAM pre_nonlin5
SELECT CASE (i2s_par_no)
CASE (1)
READ(ro, '(A)', IOSTAT = ioerr) ifgfile_path
IF (ioerr /= 0) THEN
WRITE(*, '(A)') 'Error in reading file ' // TRIM(i2s_in_file)
STOP

•                                   END IF
  

•                                   ELSEIF (ioerr.eq.0 .and. ifgfile_path.eq.'0') then^M
  

•                                       ifgfile_path=''^M
  

•                                    END IF^M
                            CASE (29)
                                    EXIT
                    END SELECT
  

@@ -214,7 +220,14 @@ PROGRAM pre_nonlin5
CYCLE
END IF

•                   READ(ro, *, IOSTAT = ioerr) spectrum_name, year, mon, day, run, lat, lon, alt, &
  

+! write(,)ro^M
+! READ(ro, , IOSTAT = ioerr) inpstr^M
+! write(,*) inpstr^M

•                    lb=fnbc(ro)^M
  

•                    lf=fbc(ro(lb:))+lb^M
  

•                    spectrum_name=ro(lb:lf-1)^M
  

+! READ(ro, *, IOSTAT = ioerr) spectrum_name, year, mon, day, run, lat, lon, alt, &^M

•                   READ(ro(lf:), *, IOSTAT = ioerr) year, mon, day, run, lat, lon, alt, &^M
                            tins, pins, hins, tout, pout, hout, sia, fvsi, wspd, wdir
                    IF (ioerr /= 0) THEN
                            WRITE(*, '(A)') 'Error in reading file ' // TRIM(i2s_in_file)
  

[ndeutscher]

fbc had to be added to the ggg_src directory and referenced in the Makefile as well. I'm still not convinced the Makefile worked correctly, but I managed to compile the three programs across several different attempts.

I've tested this with Wollongong 2014 (with full path in each entry of opus-i2s.in) and with the sodankyla test case. It seems to work. The same answers are generated for the test case, and Wollongong runs.