Accepted Solutions
Hi:
Interesting that the OP program did not work with termstr=LF, but my program DID work with termstr=LF. When I tried my simple program, to capture the length of each data line and count the number of decimals:
data explore;
infile 'c:\temp\HPA.txt' length=lg lrecl=32767 missover termstr=LF;
** read state and hold the data line;
input state $ @;
numdec = count(_infile_,'.');
putlog _n_= lg= state= numdec=;
run;here are the line lengths I got for each data line:
At least one of my data lines (for state=AR had a length of 13493, which agrees with Roger's length of one of his datalines.
I used termstr=LF in my code. So I think that is delimiting the records. I tried other values for termstr as shown below and none of these worked as well as termstr=LF:
--tried termstr=CRLF and termstr=CR neither of them did as well as termstr=LF
also tried termstr='0D0A'x and that did not work either:
But the OP has not come back to answer Tom's question about whether there are always 18 numbers after every STATE value or to comment on the number of decimal points that I found on every data line -- it seems unusual to me that there are exactly 720 periods for every data line -- that seems too regular a pattern.
And my final program
data explore;
infile 'c:\temp\HPA.txt' length=lg lrecl=32767 missover termstr=LF;
** read state and hold the data line;
input state $ @;
numdec = count(_infile_,'.');
if state gt ' ' and lg gt 0 and numdec gt 0 then do;
input val1-val720;
putlog _n_= lg= state= val1= val2= val3= numdec=;
output;
end;
run;
seemed to produce the desired SAS dataset....but required some exploration to come up with the right logic and variable counts.
cynthia
I am not sure that the file is just one row. When I open the HPA.TXT file with NoteTab Light, it looks like this:
Note how AK, AL, and AR all start at the beginning of a new line. So I wonder if this is an issue of no CR or LF at the end of a row of data or whether this was generated on some system and then FTP'd incorrectly so the CR/LF did not get translated correctly. Or something.
Anyway, I -could- read the file. There are a LOT of numbers. I just did a simple
to count the number of decimals in the INFILE buffer and each "row" for each "state" had 720 decimals, which is how I came up with the ending number for the INPUT statement.
But there is still strangeness in the file. I used this to start:
data explore;
infile 'c:\temp\HPA.txt' length=lg lrecl=32767 missover termstr=LF;
** read state and hold the data line;
input state $ @;
numdec = count(_infile_,'.');
putlog _n_= lg= state= numdec=;
run;just to get the number of decimal places (in order to read the variables after STATE) and this is what I got in the log:
So it might be that there is something weird at the end of the file, but given that STATE has the correct value for _n_ 1, 2, and 3, I think that it should be possible to read the file and just skip over the lg=1 or any lg=0 data lines that might be coming from odd control characters in the file.
This is what I tried
and the proc print looks OK, but I did not even try to show all 720 numbers.
And, of course, this is only a sample of 3, but I would be tempted to go back to the folks who designed this process and have them check what is being put at the bottom of the file. And, possibly sweet talk them into using normal CR/LF. Bottom line, is that I don't think this ASCII text file is ever going to be "readable" in Notepad in any meaningful way.
cynthia
cynthia
Just opening the sample text file you've attached with Notepad++ reveals that your lines with numbers end with a UNIX style LF
Could it be that your downstream process expects a Windows style CRLF as end of record indicator? If this process is in SAS then use TERMSTR as part of your Infile statement. http://support.sas.com/kb/14/178.html
Comment:
I'm referring to sample data HPA.txt. The larger sample file others refer to hasn't been available anymore in this thread.
Hi:
Interesting that the OP program did not work with termstr=LF, but my program DID work with termstr=LF. When I tried my simple program, to capture the length of each data line and count the number of decimals:
data explore;
infile 'c:\temp\HPA.txt' length=lg lrecl=32767 missover termstr=LF;
** read state and hold the data line;
input state $ @;
numdec = count(_infile_,'.');
putlog _n_= lg= state= numdec=;
run;here are the line lengths I got for each data line:
At least one of my data lines (for state=AR had a length of 13493, which agrees with Roger's length of one of his datalines.
I used termstr=LF in my code. So I think that is delimiting the records. I tried other values for termstr as shown below and none of these worked as well as termstr=LF:
--tried termstr=CRLF and termstr=CR neither of them did as well as termstr=LF
also tried termstr='0D0A'x and that did not work either:
But the OP has not come back to answer Tom's question about whether there are always 18 numbers after every STATE value or to comment on the number of decimal points that I found on every data line -- it seems unusual to me that there are exactly 720 periods for every data line -- that seems too regular a pattern.
And my final program
data explore;
infile 'c:\temp\HPA.txt' length=lg lrecl=32767 missover termstr=LF;
** read state and hold the data line;
input state $ @;
numdec = count(_infile_,'.');
if state gt ' ' and lg gt 0 and numdec gt 0 then do;
input val1-val720;
putlog _n_= lg= state= val1= val2= val3= numdec=;
output;
end;
run;
seemed to produce the desired SAS dataset....but required some exploration to come up with the right logic and variable counts.
cynthia
Learn how use the CAT functions in SAS to join values from multiple variables into a single value.
Find more tutorials on the SAS Users YouTube channel.
-
4892 views
-
2 likes
-
-
