How can I count letters in a text file?

No I can't. In fact, I don't really like being "Dave's Homework Helper Site" because the point of homework is for you to practice the requisite skills so you can get better at whatever you're studying. (why yes, I am a teacher)

On the other hand... :-)

This is an interesting little puzzle so what I will do is show how to write a quick, short Bourne Shell script for Linux (or Mac OS X if you crack open your Terminal.app program) that can do what you seek.

The key idea is that if you could transform the input to be one-character-per-line, it'd be unreadable for humans, but would make it really, really easy to sort and tally for a computer program.

How do you do that? With one of what I call the unsung heroes of the Unix command line, the "fold" command. Generally, people use fold to wrap overly long lines in text files (it's great for processing info prior to printing it, for example) but as with all great Unix command line utilities, it has parameters that let you change its behavior. And that's just what we'll do.

Try this yourself:

$ date | fold -w3
Wed 
Nov 
26 0
9:56
:48 
MST 
2008

That's with width=4. Turn it into "-w1" and each and every character is on its own line. (I won't reproduce it here because it's crazy long and you get the idea anyway, I hope!)

Now that each character is on its own line, it's simple to sort the output to ensure that they're in alphabetical order with "sort". To tally matching lines turns out to be a feature of another of the unsung heroes, "uniq". Check its man page and you'll see:

That's what we want, the "-c" flag. Now, to put them together:

$ date | fold -w1 | sort | uniq -c
   5  
   4 0
   2 1
   3 2
   1 5
   1 6
   1 8
   2 :
   1 M
   1 N
   1 S
   1 T
   1 W
   1 d
   1 e
   1 o
   1 v

Uh oh, there's a problem: we really want to count upper case and lower case together, so we'll need to slip one more transform into the pipeline, and we'll do it with mnemonic set names, which you might not have seen before:

tr '[:upper:]' '[:lower:]'

Put this into the pipeline and give it something more interesting to chew on (the "bash" man page) and here's what we get:

$ cat sample-input.txt | fold -w1 | \
     tr '[:upper:]' '[:lower:]' | sort | uniq -c
  13 
 570  
   3 !
   3 $
  19 '
   1 (
   1 )
  33 ,
   4 -
  19 .
   4 0
   1 1
   4 2
   1 3
   1 4
   2 5
   1 7
   2 8
   8 9
  10 :
   5 ?
 184 a
  43 b
  77 c
 102 d
 304 e
  49 f
  42 g
 105 h
 164 i
   1 j
  25 k
 122 l
  67 m
 165 n
 257 o
  55 p
   3 q
 141 r
 173 s
 259 t
  95 u
  40 v
  40 w
   3 x
  93 y
   3 z

This is a bit hard to read, so let's add one more "grep" at the end so we can just isolate the vowels and have our output readable:

$ cat sample-text.txt | fold -w1 | tr '[:upper:]' '[:lower:]' | \
     sort | uniq -c | grep -E '(a|e|i|o|u)'
 184 a
 304 e
 164 i
 257 o
  95 u

There ya go. You can easily drop this sequence of Unix / Linux commands into a shell script and use whatever file or files are specified as the input.

Notice here that as I expect, "E" is the most common vowel, followed by "O", "A", "I" and, way down the list, "U". In fact, when I feed a much larger corpus of work to this script, here's what I end up with: a (8,621), e (12,419), i (7,497), o (9,195) and u (3,660). Again, "E" is the most common, followed by "O" and fairly closely by "A". If you ever play hangman, it's very useful to know letter frequencies!