I am trying to do some basic math with files in J

Consider:

$ cat nrow
1 2 3 4 5
$ cat ncol
1
2
3
4
5

and in J:

   'b' fread 'nrow'
┌─────────┐
│1 2 3 4 5│
└─────────┘
   'b' fread 'ncol'
┌─┬─┬─┬─┬─┐
│1│2│3│4│5│
└─┴─┴─┴─┴─┘
   $".>'b' fread 'ncol'
5
   $".>'b' fread 'nrow'
1 5
   'm' fread 'ncol'
1
2
3
4
5
   'm' fread 'nrow'
1 2 3 4 5
   $".'m' fread 'nrow'
1 5
   $".'m' fread 'ncol'
5

So they're both pretty straightforwardly read in with fread, with just that difference in shape. If you're comfortable having specific file-ingestion verbs for specific arrangements of numbers in files, you could stop here.

But this might also be useful:

   nums =: [: , 0 ". 'm' fread ]
   nums 'nrow'
1 2 3 4 5
   (nums 'nrow') -: nums 'ncol'
1

With , (ravel) this just gives you numbers from a file, however they're arranged in the file, and replacing any non-number with 0 (instead of evaluating it)

   'a 1 2 b 3' fwrite 'arow'
9
   nums 'arow'
0 1 2 0 3

The first issue to deal with is the line endings. Using freads from the standard library will read your text file and ensure that each line is terminated by LF, no matter whether lines were originally CRLF, CR or just LF terminated. Use fread, with no left argument, if you want to read the bytes exactly as they were in the file.

Let's use the following space-delimited file (numbers.txt) as an example:

1 -3.93 17 -7 564
2 4.27 12 3 234
3 -1.90 22 5 728
4 0.00 10 -4 442

You can read the file as follows:

   0 ". 'm' freads 'numbers.txt'
1 _3.93 17 _7 564
2  4.27 12  3 234
3  _1.9 22  5 728
4     0 10 _4 442

Under the covers the 'm' left argument to freads is doing this:

   0 ". ];._2 freads 'numbers.txt'

The ;._2 says to use the last byte in the file (which is now LF) as the delimiter, remove them from the result and apply the verb to its left (] in this case) to each line. The result is a character array with a row for each line and columns corresponding to the longest line in the file (shorter lines are padded with spaces). 0 ". is then attempts to parse the array as numbers, where it can't it will replace the input with 0 (I often use _999 so it is more obvious that something hasn't converted correctly).

If the file is delimited by something other than a space, then you need to parse the fields otherwise you'll get interesting results:

   0 ". ];._2 freads 'numbers.csv'
0 24.2712 0 0

If it is a simple file you could try the one of following:

   0 ". ];._2 ', ' charsub freads 'numbers.csv'  NB. replace delimiter with spaces
1 _3.93 17 _7 564
2  4.27 12  3 234
3  _1.9 22  5 728
4     0 10 _4 442

   0 ". > ','&cut;._2 freads 'numbers.csv'  NB. box fields in each line
1 _3.93 17 _7 564
2  4.27 12  3 234
3  _1.9 22  5 728
4     0 10 _4 442

If it is more complicated (quoted delimiters etc) or if you just find it easier, you can use the tables/csv or tables/dsv addons:

   load 'tables/csv'
   0 ". > readcsv 'numbers.csv'
1 _3.93 17 _7 564
2  4.27 12  3 234
3  _1.9 22  5 728
4     0 10 _4 442

   0 ". > ',' readdsv 'numbers.csv'
1 _3.93 17 _7 564
2  4.27 12  3 234
3  _1.9 22  5 728
4     0 10 _4 442

So let's take it from the point that you have read your file into the J environment.

txt=: 1!:1 < F

In txt, if a number is on a different line than another then there must be a CR (carriage return) or LF (linefeed)character between the characters (or maybe CRLF juxtaposed depending on the way linefeeds are represented). So, if we replace those LF and CR characters with ' '(blank space) using } (Amend adverb)

txt1=: ' '  I.@: ( (CR , LF)  e.~ ]) } txt

then the multi-line string becomes one line. Convert that line to numbers using 0 ". and you can manipulate your numbers as you wish.

If you knew that each line either began or ended with a blank space then you could simply remove LF and CR using -.

txt1=: txt -. LF , CR

but this would run numbers together across lines if there were not blanks to separate them.