sort a list numerically and alphabetically

Question

I'm trying to find most common word sorted alphabetically. For example:

  2 went
  2 wonder
  2 wont
  3 began
  3 little
  3 moment
  3 rabbit
  3 thing
  3 till
  4 alice
  4 bottle
  4 came
  4 sure
  4 window

The output should be alice (has highest value and it is the first word alphabetically).

I'm doing sort -nr and can't figure out what to do next.

Answer 1

To sort by two fields, you need to tell sort what they are, and how to sort them; for example:

sort -k1,1nr -k2b < input

sorts by field 1 (-k1,1) numerically in reverse (descending) order; for lines where field 1 is equal, secondarily sort by the rest of the line (-k2) normally (lexically) not including the leading blanks (the spacing between the first and second field) in the sort key (b). The output on your sample input is:

  4 alice
  4 bottle
  4 came
  4 sure
  4 window
  3 began
  3 little
  3 moment
  3 rabbit
  3 thing
  3 till
  2 went
  2 wonder
  2 wont

Answer 2

Though Jeff already answered your question using only sort, I thought I'd post my approach anyway. I didn't know about sort's ability to sort by two fields, so I used it merely for "pre"-sorting you input. Since the words are ordered Z-A in every section, we just need to reverse their order. For this, I'm using awk:

sort -r file | awk '{if($1==x){var=$0"\n"var} else{print var;var=$0;x=$1}} END{print var}' | grep -v ^$

The grep -v ^$ is only there to remove the first line of awk's output, which is empty.