topic I/O Challenges with FORTRAN in Intel® Fortran Compiler

I/O Challenges with FORTRAN

nightghost — Wed, 15 Dec 2010 12:05:36 GMT

Good morning to everyone, I have some question about the use of the READ statement when I use it to read from a file.
I have an input file, which contains the following information:

10 ! Index of a 3D array, containing some information let's say the number of a composite laminate

6 ! Number of plies
[90/45/0]s! A string used for a subroutine, which creates a selection menu.
90.
45.
0.
0.
45.
90. ! The number of plyes orientations in degree.

The latter block is repeated 10 times until EOF is found.

I wrote the following subroutine, but I got a runtime error everytime with INTEL FORTRAN.

This code was running without problems on a VAX machine many years ago when I was at University.

Thanks for the help.

Gianfranco

C-------------------------------------------------------------------------------

SUBROUTINE CHOLAM(N,THEAT,LAMCHO)

C-------------------------------------------------------------------------------

DOUBLE PRECISION THEAT(40),THEATA(20,40)

CHARACTER*30 LAMCHO,LAMTYP(40)

INTEGER CHOICE,NPLYS(20),NLAMS

OPEN(UNIT=100,FILE='test.dat',ACCESS='SEQUENTIAL')

READ(50,30)NLAMS

WRITE(100,30)NLAMS

DO 200 J=1,NLAMS

READ(50,30)NPLYS(J)

30 FORMAT(I2/)

DO 100 I=1,NPLYS(J)+1

IF(I.EQ.1) THEN

READ(50,10)LAMTYP(J)

10 FORMAT(A30/)

ELSE

READ(50,20)THEATA(J,I-1)

20 FORMAT(F6.2/)

ENDIF

100 CONTINUE

200 CONTINUE

WRITE(*,*)"Enter # of laminate/ply choice..."

DO 300 J=1,NLAMS

WRITE(*,250)J,LAMTYP(J)

250 FORMAT(I3,': ',A30)

300 CONTINUE

READ(*,*)CHOICE

N=NPLYS(CHOICE)

LAMCHO=LAMTYP(CHOICE)

DO 400 K=1,N

THEAT(K)=THEATA(CHOICE,K)

400 CONTINUE

CLOSE(100)

RETURN

END

I/O Challenges with FORTRAN

Arjen_Markus — Wed, 15 Dec 2010 12:24:30 GMT

What is the error message that you get?

Without the actual input file it is impossible to reproduce the error - adding that might help as well
to diagnose the problem.

Regards,

Arjen

I/O Challenges with FORTRAN

Steven_L_Intel1 — Wed, 15 Dec 2010 15:56:21 GMT

Yes, the error message would help, as would the actual input file as an attachment because posting it as text in the forum removes leading blanks. Two things come to mind immediately. The first is that all of your formats end in /, indicating that a record should be skipped. This makes sense for the initial read, which seems to have a blank line following it, but not for the subsequent reads. Also the read of the number of plies is I2, but you show only one digit. I would expect this to work with Fortran 90 and later compilers, which assume BLANK='NULL', but VAX Fortran, being a F77 compiler, would read "6 " as 60.

I/O Challenges with FORTRAN

jimdempseyatthecove — Thu, 02 Jun 2011 20:08:35 GMT

Review your lam.txt file. It appears to be in error.

The format seem to follow

n
[text]
data-1
data-2
...
data-n

m
[other text]
data-1
...
data-m

However, in reading the file, it does not follow this pattern

GIGO

Jim Dempsey

I/O Challenges with FORTRAN

Steven_L_Intel1 — Thu, 02 Jun 2011 20:37:18 GMT

The problem appears to relate to the use of a trailing slash in each of the formats. This causes one additional input record to be skipped. If this "worked' in VAX Fortran then that compiler made a mistake, though I don't recall any such issues (I was the project lead for that compiler for many years.)

However, Jim is also correct that the data file doesn't follow the pattern.

Here's a version of the code fwith the slash problem fixed.

[fortran]DOUBLE PRECISION THEAT(40),THEATA(20,40)
CHARACTER*30 LAMCHO,LAMTYP(40)
INTEGER CHOICE,NPLYS(20),NLAMS

OPEN(UNIT=100,FILE='test.dat',ACCESS='SEQUENTIAL')
READ(50,*)NLAMS
WRITE(100,30)NLAMS

DO 200 J=1,NLAMS
READ(50,30)NPLYS(J)
30 FORMAT(/I2)
DO 100 I=1,NPLYS(J)+1
IF(I.EQ.1) THEN
READ(50,10)LAMTYP(J)
10 FORMAT(A30)
ELSE
READ(50,20)THEATA(J,I-1)
20 FORMAT(F6.2)
ENDIF
100 CONTINUE
200 CONTINUE

WRITE(*,*)"Enter # of laminate/ply choice..."
DO 300 J=1,NLAMS
WRITE(*,250)J,LAMTYP(J)
250 FORMAT(I3,': ',A30)
300 CONTINUE

READ(*,*)CHOICE

N=NPLYS(CHOICE)
LAMCHO=LAMTYP(CHOICE)


DO 400 K=1,N
THEAT(K)=THEATA(CHOICE,K)
400 CONTINUE


CLOSE(100)

END[/fortran]

I/O Challenges with FORTRAN

jimdempseyatthecove — Thu, 02 Jun 2011 23:35:59 GMT

In the lam.txt file the 1st group marked with 6 data points has 5 data points, same with the 3rd group. The 4th group of 6 has 6 data points but no empty line following the group (what you indicate is the problem). So either there is missing data points or the count is incorrect. I am not assuming that reading a blank line is intended to be equivilent to reading a 0.

Jim Dempsey

I/O Challenges with FORTRAN

nightghost — Tue, 07 Jun 2011 09:36:35 GMT

Hi Steve, thanks for your help. I finally fixed the bug and after a step by step debug I found also that the main reason wasn't only the erroneous FORMAT cards, but also the fact that the files were UNIX format. When I realized it I finally found that the code is working nice. I also had to remove the blank lines in order to keep the code as simple as possible. Blanks line are nice for the user to understand the different blocks, but they need more coding to be properly handled and for now the porting is the main task of my current job. We don't want to re-invent the wheel, do we?

Anyway thanks for your time and assistance.

Cheers,

Gianfranco

I/O Challenges with FORTRAN

nightghost — Tue, 07 Jun 2011 09:40:29 GMT

Hi Jim,

thanks for having spotted the mistake in the file. I later realized that the blank lines weren't existing in theoriginal input file. One of my students added for a better readability, but the code wasn't written to handle them. I was lucky I had the possibility to restore an old backup and inspect the files. Moreover, the file was in UNIX format instead ofa DOS one.

Cheers,

Gianfranco Iasiello

I/O Challenges with FORTRAN

jimdempseyatthecove — Tue, 07 Jun 2011 14:59:10 GMT

>>One of my students added for a better readability,

Maybe they had an adversion to the number 5 too and replaced it with a 6 ;)

The blank line parsing and/or embedded comment parsing is relatively easy to code. If it makes reading your input file easier then consider adding it back. Data format becomes:

[blank] (any number of blank lines)
[! comment] (comingled with any number of comments)
[C comment if you wish]
n
"[your tag here]" (required tag)
data-1
...
data-n
((repeat above until done))

You only need one section to parse the blank and comment lines and that is where you obtain the datacount.

Also note that you can code your data as

n
"[your tag here]"
data-1 ! X (embed comment with data)
data-2 ! Y

The comments will slow down the input read a bit but I would think you would only comment parameter files and not a multi-gigabyte input stream.

Jim Dempsey

Hi, I need help in figuring

Suresh_A_ — Mon, 03 Nov 2014 03:45:21 GMT

Direct Access Formatted Read Error

Hi, I need help in figuring out why my read statement for formatted direct access statement does not work with Intel Fortran. It worked without any problem with Visual Fortran 2005 and Compaq Fortran.

The program listing is as follows:

1       Program TestLPIntfDll
      2
      3 C
      4 c     INCLUDE      'GEPARM.FOR'
      5    INCLUDE      'GEBLK1.FOR'
      6 c     INCLUDE      'GEBLK2.FOR'
      7
     14       CHARACTER(LEN=100) :: FILEPATH
     15       CHARACTER(LEN=80) ::CBLOCK
     16    CHARACTER(LEN=20) :: FLNAME
     17       INTEGER*2 :: NROWS, NCOLS
     18
     19
     20       FLNAME="LPTEST1.DAT"
     21
     22       OFFLUN=20
     23       DBGLUN=25
     24       DATPRT=30
     25       LUNPRT=35
     26       IRDWR=1
     27
     28       FILEPATH="G:\oms-Development\Version-VB6\Interfaces\" //
     29      1 "omsLPIntf\Input\LPTEST1.DAT"
     30
     31
     32       OPEN (UNIT=20, FILE=FILEPATH, STATUS='OLD',
     33    1 ACCESS='DIRECT',RECL=80, FORM='FORMATTED',
     34    2 ERR=80, IOSTAT=IOS)
     35
     36       Open (UNIT=25,file='C:\OMSLLC\DEBUG-' //
     37      1 FLNAME,STATUS='REPLACE')
     38
     39       Open (UNIT=30,file='C:\OMSLLC\INPUT-DATA-' //
     40      1 FLNAME,STATUS='REPLACE')
     41
     42       Open (UNIT=35,file='C:\OMSLLC\LP-MATRIX-SOLN-' //
     43      1 FLNAME,STATUS='REPLACE')
     44
     45    WRITE (25,800) FILEPATH
     46 800   FORMAT(A80)
     47
     48 C      0. LP PROBLEM RUN IDENTIFICATION TITLE
     49 *
     50       DO 5 IRC = 1,3
     51       READ (20,FMT=990,REC=IRC,ERR=80) TITLE(IRC)
     52    WRITE (30,FMT=990,ADVANCE='NO') TITLE(IRC)
     53 5     CONTINUE
     54 *
     55       DO 51 IRC = 4,6
     56       READ (20,FMT=990,REC=IRC,ERR=80) CBLOCK
     57    WRITE (30,FMT=990,ADVANCE='NO') CBLOCK
     58 51    CONTINUE
     59
     60 !6     IRC = 6
     61 *
     62 *---- 1. GENERAL FIXED LP PARAMETERS
     63 *

Page 2 Source Listing TESTLPINTFDLL
2014-11-02 20:43 TestLPIntfDll.FOR

     64 C      IRC = IRC +1
     65       READ (20, FMT=1000,REC=IRC,ERR=80)    NROWS
     66    READ (20, FMT=1000,REC=IRC+1,ERR=80) NCOLS
     67       READ (20, FMT=1000,REC=IRC+2,ERR=80) NFXMCF
     68 C
     69    WRITE (25,6002) NROWS, NCOLS, NFXMCF
     70
     71 80   CONTINUE
     72       WRITE (25,1050,ADVANCE='NO') FLNAME, IRC
     73 *
     74 *---- FORMAT STATTEMENTS
     75 *
     76 990   FORMAT (A72)
     77 1000 FORMAT (56X,I3)
     78 1050 FORMAT (/1X,' GEDATA ERROR - IN READING FILE ',A20,' AT RECORD',
     79      1' NUMBER ',I3)
     80 6002 FORMAT (2X,3I3)
     81
     82       CLOSE(20)
     83       CLOSE(25)
     84       CLOSE(30)
     85       CLOSE(35)
     86
     87       !CLOSE(OFFLUN)
     88       !CLOSE(DBGLUN)
     89       !CLOSE(LUNPRT)
     90
     91       END

The input file is fixed columns data as follows:

*TITLE                      LP TEST CASE -1
*TITLE          SMALL REFINERY WITH REFORMER AND CRACKER
*TITLE
*
****************************** FIXED LP DATA ***************************
*
NUMBER OF ROWS                             =              3
NUMBER OF VARIABLES                        =              5
NUMBER OF FIXED MATRIX COEFFS              =              9
*** ROW NAMES AND RIGHT HAND SIDES ***
³ROWNAM³    ³-- RHS --³
PIPESTIL      1000.00
REFORMER       480.00
CRACKER        640.00
*** VARIABLE NAMES, LOWER / UPPER BOUNDS AND COSTS ***
³VARNAM³    ³ LOW BNDS ³³ UPP BNDS ³³- COST -³³- SOLN -³³VARSTA³
LIGHTCRD         0.00       1000.00     -0.6000      700.00       2
HEAVYCRD         0.00       1000.00     -0.5000      300.00       2
PSSLACK          0.00       1000.00      0.0000        0.00       1
RFSLACK          0.00        480.00      0.0000        0.00       1
CRSLACK          0.00        640.00      0.0000        0.00       1
*** FIXED MATRIX COEFFICIENTS ***
³ROWNAM³    ³VARNAM³     ³FIXD COEFF³
PIPESTIL    LIGHTCRD           1.0000
PIPESTIL    HEAVYCRD           1.0000
PIPESTIL    PSSLACK            1.0000
REFORMER    LIGHTCRD           0.6000
REFORMER    HEAVYCRD           0.2000
REFORMER    RFSLACK            1.0000
CRACKER     LIGHTCRD           0.4000
CRACKER     HEAVYCRD           0.8000
CRACKER     CRSLACK            1.0000
*** END OF FILE ***

and I get error for record no 7 as follows:

G:\oms-Development\Version-VB6\Interfaces\omsLPIntf\Input\LPTEST1.DAT

GEDATA ERROR - IN READING FILE LPTEST1.DAT AT RECORD NUMBER 7

I can not figure out. I am sure it must be simple as I may have not specified some new option or parameter in the formatted read statement.

Any suggestion would be appreciated.

Thanks

The following items prevent

mecej4 — Mon, 03 Nov 2014 12:25:00 GMT

The following items prevent one from running the program to make a diagnosis: 1) the include file(s) are not provided; 2) the program source and, even more importantly, the direct access data file, were posted as text inside the message; please provide files as attachments combined into a zip-file, and state the compiler version, platform, and compiler options used.

You read 3 lines for the

Les_Neilson — Mon, 03 Nov 2014 12:50:00 GMT

You read 3 lines for the title.

You read 3 lines for CBLOCK

You assign 6 to IRC

You have commented out the increment of IRC (line 64)

Therefore:

You are reading record number 6 for IROWS. Which according to my eyes and counting the lines in the input file is an asterix. Which means NROWS will be zero.

You then read NCOLS (record 7) and then NFXMCF (record 8)

According to the code the format for reading number of rows, columns and matrix coefficients is (56x,i3)
Are you sure that there are 53 characters and 2 spaces before the numbers? (I haven't counted.)

Are there tabs in the data file?

<edit>

Another thought. You specify a record length of 80 - presumably bytes. Could your settings be: record length is specified in words ( a record would then be 80 * word length)?

</edit>

Les

Quote:You are reading record

mecej4 — Mon, 03 Nov 2014 13:41:00 GMT

You are reading record number 6 for IROWS.

Not quite. The error is not in the value of IROWS, if you note that the code uses the property of DO loops that the index variable has the value that it had when the the DO iteration test failed -- in this case, 7. The problem is that the code reads a direct access file, and traps I/O errors by including ERR= and IOSTAT= clauses, but does nothing to check the errors returned, nor is the data file itself provided. The representation used for direct access files can vary from compiler to compiler on an operating system that does not have native record files.

Check Les's suggestion about

jimdempseyatthecove — Mon, 03 Nov 2014 14:00:03 GMT

Check Les's suggestion about hidden tab characters.

Some of "modern" editors of the last 20 years have (had) the nasty habit of inserting Tab characters for indentation thus making the columns as viewed on screen not at the actual column as read by READ.

If your editor has an option to view tabs, then turn it on when editing Fortran formatted data files.

Jim Dempsey

Quote:mecej4 wrote:

Les_Neilson — Mon, 03 Nov 2014 14:04:43 GMT

mecej4 wrote:

Quote:

You are reading record number 6 for IROWS.

Not quite. The error is not in the value of IROWS, if you note that the code uses the property of DO loops that the index variable has the value that it had when the the DO iteration test failed -- in this case, 7. The problem is that the code reads a direct access file, and traps I/O errors by including ERR= and IOSTAT= clauses, but does nothing to check the errors returned, nor is the data file itself provided. The representation used for direct access files can vary from compiler to compiler on an operating system that does not have native record files.

Ah you are correct, I totally missed the "!" at the beginning of line 60 (Other comments use "C")

Les

I suggest, as a diagnostic

Steven_L_Intel1 — Mon, 03 Nov 2014 14:14:55 GMT

I suggest, as a diagnostic aid, temporarily removing the ERR= clauses from the READ statements and seeing what the complete error message is. I will also comment that providing a source listing file is not conducive to analysis, as it makes it difficult for us to build and run the program. (Of course it doesn't help that the include files are not provided.) For us to best help you, please provide a ZIP containing all the source files, the project and solution files, and the data file(s) so that we can build and run the program.

I am not sure what you mean by "Visual Fortran 2005", but I'll assume you mean some version of Intel Visual Fortran working in Visual Studio 2005.

Hi All,

Suresh_A_ — Tue, 04 Nov 2014 05:27:54 GMT

Hi All,

I have enclosed all files related to this problem. I tried Steve's suggestion of removing ERR=80 and it gave the same message. The pgm reads 1-6 records as A80 and it is OK. It gives problem only in record=> where it supposed to skip 56 Characters and then read number in I3 format in columns 57-59.

The data file was created in fixed column format without using tabs.

I appreciate all your help in figuring out the problem with this File I/O program.

Also, it worked in Compaq fortran in VS 2005. Now, I am upgrading both compiler and the software system, it errors out.

I am sure in the end it may be a simple fix.

Thanks again

Sorry I missed few files in

Suresh_A_ — Tue, 04 Nov 2014 05:31:49 GMT

Sorry I missed few files in the zip. Here is included all files.

In your project file you have

Les_Neilson — Tue, 04 Nov 2014 09:27:00 GMT

In your project file you have VMS Compatibility set True.

From the help:

" The -vms option affects the record length for direct access and relative organization files.
The buffer size is increased by 1 to accommodate the deleted record character."

Remove this option.

If this is the actual program you are working with then I would suggest that you do not need direct access for the data file at all.

Just read the data sequentially.

Les

Les,

Suresh_A_ — Fri, 07 Nov 2014 02:52:30 GMT

Les,

Thanks, It worked.

However VMS option was selected to use /LIST option in the include statement and that VMS option nullified the direct access file record length.

So, I moved to sequential format and it worked. There was actually no reason to use direct access and whatever it was could be accomplished with sequential file access method.

Thanks Again for your help

Suresh