Ruby command line fu

ruby-command-line-fu.md

Ruby in place of sed / awk

One of Ruby's major influences was Perl, leaving well equipped to replace bash, sed, and awk. Lets take a look at how to do that.

The -e flag

When you're composing a light-weight 1 off command line invocation, you don't want to have to write your program in a file. The -e flag tells Ruby that the next argument is your program.

$ ruby -e 'puts "hello world"'
hello world

The -n and -p flags, and $_ variable

To replace sed and awk, we're going to need to iterate over every line of input. The -n and -p flags do this. As they iterate, they place the current line in the variable $_, then run your program. The difference is that -p will then print $_.

$ echo -e "a\nb\nc" | ruby -n -e 'puts $_.upcase'
A
B
C

$ echo -e "a\nb\nc" | ruby -p -e '$_ = $_.upcase'
A
B
C

Exercise: Write grep

Given a markdown file on stdin, print its headers. Assume they're using hash style headers, not equal / dash style.

Here is an example input file:

$ curl -sL 'https://raw.githubusercontent.com/ruby/ruby/trunk/README.md'

Here is my solution, we'll soon learn ways to improve on it!

$ curl -sL 'https://raw.githubusercontent.com/ruby/ruby/trunk/README.md' | ruby -n -e 'puts $_ if $_[/^#+/]'
# What's Ruby
## Features of Ruby
## How to get Ruby
## Ruby home page
## Mailing list
## How to compile and install
## Copying
## Feedback
## Contributing
## The Author

Cache command line tool

So, the above might have been a bit painful since it can take over a second for the curl to complete. I'm often in situations like that, where I need to experiment with what to do next after running some expensive command.

That means I'll have to pay that expense a lot of times, which makes it cumbersome to do the exploration. To avoid that expense, I wrote a tool called cache, which runs a command line invocation, records its stdout/stderr/exit status, and then replays them in future invocations. Once I've got the command-line invocation figured out, I just remove the call to cache. You can get its source code here. Then put that somewhere in your path and chmod +x it. I usually put new / experimental command line tools in ~/bin, which I've added to my path in bash, and fish.

You can use it like this:

$ time cache ruby -e 'sleep 1; puts "hello"'
hello
        1.31 real         0.16 user         0.05 sys

$ time cache ruby -e 'sleep 1; puts "hello"'
hello
        0.08 real         0.06 user         0.01 sys

So the way I dealt with the previous exercise was like this:

$ cache curl -sL 'https://raw.githubusercontent.com/ruby/ruby/trunk/README.md' \
>  | ruby -n -e 'puts $_ if $_[/^#+/]'

And then I removed the call to cache once I'd gotten it figured out.

Join flags to shorten the invocation

When a short flag doesn't take args, you can join it with the next short flag, here we do -ne instead of -n -e

$ echo -e "a\nb\nc" | ruby -ne 'puts $_.upcase'
A
B
C

print without args, prints $_

echo -e "a\nb\nc" | ruby -ne 'print; print;'
a
a
b
b
c
c

Regex in a conditional matches against $_

echo -e "a\nb\nc" | ruby -ne 'print if /[ab]/'
a
b

Improving our grep

Using these pieces, we can improve our grep!

# before
$ ruby -n -e 'puts $_ if $_[/^#+/]'

# after
$ ruby -ne 'print if /^#+/'

Shorthand interpolation

When a variable begins with a sigil, eg our cash symbol in $_, you can omit the curly braces from the interpolation.

$_ = "b"     # => "b"
"a #{$_} c"  # => "a b c"
"a #$_ c"    # => "a b c"

This does require that the next letter after its name not be a valid part of a name. Eg $_c is a valid variable name, so we couldn't use the shorthand interpolation for "a#{$_}c".

The current line number: $.

$ echo -e "a\nb\nc" | ruby -ne 'puts $., $_'
1
a
2
b
3
c

Exercise: Add line numbers to the input

Here is my solution:

$ echo -e "a\nb\nc" | ruby -ne 'puts "#$.\t#$_"'
1       a
2       b
3       c

Regexp#~@ matches the regexp against stdin

This is useful when Ruby isn't quite smart enough to figure out that the regex is being used in a conditional.

$ ruby -ne 'print if /^#+/'
a
b

The -l flag performs newline processing

It will chomp newlines from $_

$ echo -e "a\nb\nc" | ruby -ne 'p $_'
"a\n"
"b\n"
"c\n"

$ echo -e "a\nb\nc" | ruby -lne 'p $_'
"a"
"b"
"c"

It also causes print to behave like puts

$ ruby -e 'print "a"; print "b"'
ab

$ ruby -le 'print "a"; print "b"'
a
b

Exercise: Append line length to nonempty lines

Here is an example input:

$ curl -sL 'https://raw.githubusercontent.com/ruby/ruby/trunk/BSDL'

Here is my solution:

$ curl -sL 'https://raw.githubusercontent.com/ruby/ruby/trunk/BSDL' | ruby -lne 'print $_, ~/./ && " (#{$_.length})"'
Copyright (C) 1993-2013 Yukihiro Matsumoto. All rights reserved. (64)

Redistribution and use in source and binary forms, with or without (66)
modification, are permitted provided that the following conditions (66)
are met: (8)
...

The -a flag splits the input and saves it in $F

$ echo -e "1 a\n22 bb\n333 ccc" | ruby -ane 'p $F'
["1", "a"]
["22", "bb"]
["333", "ccc"]

The -F flag sets the input field separator

This will affect how the the -a flag splits the records. Here, we'll use it to print the first three usernames from /etc/passwd

$ cat /etc/passwd | ruby -ne 'print unless /^#/' | ruby -F: -ane 'puts $F[0] if $. <= 3'
nobody
root
daemon

Dir globbing is a fast way to find files

The docs are here, Especially useful are single and double splats. Here, we use the single splat to find all the C source files at the root directory of Ruby, we then print the first 3

$ ruby -e 'puts Dir["*.c"].take 3'
addr2line.c
array.c
bignum.c

Backticks execute a system command and return its output

$ ruby -e 'p `whoami`'
"josh\n"

Exercise: Find ruby files in the standard library

Output the first 5 files, alphabetically, that are anywhere in the standard library.

• You can find the stdlib's location with gem which csv.
• You can get the directory of a file with File.dirname(filepath).
• You'll need to use the double splat for a recursive glob

Here is my solution:

$ ruby -e 'puts Dir["#{File.dirname `gem which csv`}/**/*.rb"].sort.take(5)'
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/English.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/abbrev.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/base64.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/benchmark.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/bigdecimal/jacobian.rb

You can change directories with Dir.chdir

Also note that the block form will cd you back when done executing.

$ ruby -e 'p Dir.pwd; Dir.chdir("/"); p Dir.pwd'
"/Users/josh"
"/"

$ ruby -e 'p Dir.pwd; Dir.chdir("/") { p Dir.pwd}; p Dir.pwd'
"/Users/josh"
"/"
"/Users/josh"

Exercise: Same as before, but with relative paths

ruby -e 'Dir.chdir File.dirname `gem which csv`; puts Dir["**/*.rb"].sort.take(5)'
English.rb
abbrev.rb
base64.rb
benchmark.rb
bigdecimal/jacobian.rb

Exercise: Print the stdlib's 10 shortest files and their sizes

• It's much quicker to use File.stat(filename).size than File.read(filename).size, though they're not completely equivalent, but close enough for most uses.
• You'll probably want to use more than one Ruby program
• Note that sort -n will sort numerically instead of alphabetically

Here's my solution:

$ ruby -e 'puts Dir["#{File.dirname `gem which csv`}**/*.rb"]' \
>  | ruby -lne 'print File.stat($_).size, "\t", $_' \
>  | sort -n \
>  | head
50	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/drb.rb
59	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/optionparser.rb
85	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/tkfont.rb
85	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/tktext.rb
88	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/tkafter.rb
88	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/tkentry.rb
91	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/tkcanvas.rb
91	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/tkdialog.rb
91	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/tkmacpkg.rb
91	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/tkwinpkg.rb

Flipflops are useful in conditionals

A flipflop looks like a range, but it's syntactically placed where a conditional is. Its endpoints are boolean expressions. It evaluates to true when it's flipped on, and false when it's flipped off.

It flips on when the LHS evaluates to true, and flips off when the RHS evaluates to true.

$ echo -e "a\nb\nc\nd\ne" | ruby -ne 'print if ($.==2)..($.==4)'
b
c
d

Ranges in conditionals are actually flip flops, they match against $., so we can simplify the above with:

echo -e "a\nb\nc\nd\ne" | ruby -ne 'print if 2..4'
b
c
d

Regexes as flipflop endpoints match against stdin.

Exercise: Print Ruby's description of -n from the manual

• You can normalize the output of man by piping it through col -b
• Use a regexp flipflop, because it will be more resilient to change than line numbers

Here's my solution:

$ man ruby | col -b | ruby -ne 'print if /^\s*-n/../end/'
        -n      Causes Ruby to assume the following loop around your
                script, which makes it iterate over file name arguments
                somewhat like sed -n or awk.

                        while gets
                          ...
                        end

gsub is added to main and operates on $_

There are several other methods like this, but gsub is the most useful one.

$ echo -e "cat\ncot\ncut" | ruby -pe 'gsub "c", "p"'
pat
pot
put

The -n and -p flags iterate over ARGF

So it's not that these flags iterate over lines from stdin. Rather, they iterate over lines from the program's input. It is ARGF's job to figure out what that means. If it sees no arguments in ARGV, then it will iterate over stdin. But if it does see arguments, then it will assume they are filenames and iterate over them as files.

This is useful with -i to do mass renamings in your project (and git so you don't have to fear fucking it up).

Can also be useful with $<.filename

Exercise: Find the 10 most required files in the stdlib

• Use xargs to place the input lines into ARGV for the Ruby script

$ ruby -e 'puts Dir["#{File.dirname `gem which csv`}/**/*.rb"]' \
>  | xargs ruby -ane 'puts $F[1] if /^\s*require /' \
>  | ruby -lpe $'gsub /["\']/, ""' \
>  | sort \
>  | uniq -c \
>  | sort -nr \
>  | head
 289 tk
  53 tkextlib/iwidgets.rb
  46 tkextlib/setup.rb
  35 tkextlib/tcllib.rb
  35 rubygems/command
  34 tkextlib/bwidget.rb
  27 rubygems
  26 thread
  26 fileutils
  24 tkextlib/blt.rb

BEGIN/END run before/after the script

All BEGIN blocks will run before the first line, END will run before the program exits. They are scoped to main, so you can use them to initialize variables.

Here, we'll use them to reverse a script.

$ echo -e "1\n2\n3" | ruby -ne '
>  BEGIN { lines = [] }
>  END   { puts lines.reverse }
>  lines << $_'
3
2
1

Exercise: Print the number of lines of input

Here's my solution:

$ ruby -ne 'END { p $. }'

Exercise: Find the files in the stdlib with the most requires

$ ruby -e 'puts Dir["#{File.dirname `gem which csv`}/**/*.rb"]' | \
>  xargs ruby -r pp -ne $'
>  BEGIN { fns = Hash.new 0 }
>  fns[$<.filename] += 1 if /^\s*require /
>  END {
>    puts fns.sort_by(&:last).reverse.take(10).map { |f, n| "#{n}\t#{f}" }
>  }'
32	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/rubygems/resolver.rb
25	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/rubygems.rb
25	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/rubygems/test_case.rb
18	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/psych.rb
17	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/webrick.rb
14	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/rexml/document.rb
14	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/net/http.rb
14	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/rubygems/package.rb
13	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/rubygems/specification.rb
13	/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/rss/maker.rb

Exercise: Find the files in the stdlib that require "date"

$ ruby -e 'puts Dir["#{File.dirname `gem which csv`}/**/*.rb"]' | \
>  xargs ruby -rpp -ane $'
>   BEGIN { fns = Hash.new { |h, k| h[k] = [] } }
>   fns[$F[1].gsub /[\'"]/, ""] << $<.filename if $F[0] == "require"
>   END { puts fns["date"] }
>  '
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/csv.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/json/add/date.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/json/add/date_time.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/optparse/date.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/psych/deprecated.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/psych/scalar_scanner.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/psych/visitors/to_ruby.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/time.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/xmlrpc/create.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/xmlrpc/datetime.rb
/Users/josh/.rubies/ruby-2.3.3/lib/ruby/2.3.0/xmlrpc/parser.rb

Exercise: Write head

You can use ruby -e 'puts *?a..?z' to print the letters "a" through "z", pipe them through a program that filters to the first 10 letters ("a" through "j")

# one solution
$ ruby -e 'puts *?a..?z' | ruby -ne 'print if 1..10'

# another solution
$ ruby -e 'puts *?a..?z' | ruby -ne 'print if $. <= 10'

# another solution
$ ruby -e 'puts *?a..?z' | ruby -pe 'exit if 10 < $.'

Exercise: Write tail

Same as above, it should print "q" through "z".

$ ruby -e 'puts *?a..?z' | ruby -ne 'BEGIN { lines = [] }; lines << $_; END { puts lines.last 10 }'

Exercise: Write cat

Replace cat in the script below with your program, it should have the same output.

cat `gem which json/version`

Here's my solution:

$ ruby -pe '' `gem which json/version`

Exercise: Find the gem with the longest name

You can get the list of all gem names with gem list --all --remote --no-versions

Here's my solution:

$ gem list --all --remote --no-versions | ruby -e 'puts $stdin.readlines.sort_by { |g| -g.length }.take(10)'
ivyxxcspcqlaocvjbghawvbdartwsfffurhnqzlwvsbgieweawfntuwecdcminmiaunqteqgbrfuxppntjdvyvsswxwepnbfqstnrnsotrhndihkudyahthaxatviwrwtgllwbqhibouqctrxtypac
XHg4NFx4QzdceDE4cFx4QzRceEM5XHhGRVx4MDBceEVDXHg5Q1x4RUZceEI5XHhDMGlceEFFfkNOXGVcdlx4OTNceEE5XHhDNw
rabbit-slide-pmq20-tracing-a-memory-leak-in-a-long-running-eventmachine-application
xlmydsykwnrfbnvjffqcokoorkbskzzhrtgnzxkapmjtffjfkwcvwklmsrzwfiatwigrvmftpbybbeqi
rabbit-slide-kou-readable-code-workshop-for-pioneer-share-readable-code
multi_json_super_best_fun_with_streaming_amazing_its_a_spoon_not_a_fork
rabbit-slide-kou-mysql-and-postgresql-and-japanese-full-text-search-2
rabbit-slide-kou-mysql-and-postgresql-and-japanese-full-text-search-3
computer_please_do_you_happen_to_know_the_time_please_thank_you_meow
rabbit-slide-kou-mysql-and-postgresql-and-japanese-full-text-search

If you only needed the longest line, you could make this much more efficient by tracking the longest one you'd seen so far, as you iterate over every line. Give that one a try (use BEGIN and END):

$ cache gem list --all --remote --no-versions \
>  | ruby -lne '
>     BEGIN { longest = "" }
>     END   { puts longest }
>     longest = $_ if longest.length < $_.length
>    '
ivyxxcspcqlaocvjbghawvbdartwsfffurhnqzlwvsbgieweawfntuwecdcminmiaunqteqgbrfuxppntjdvyvsswxwepnbfqstnrnsotrhndihkudyahthaxatviwrwtgllwbqhibouqctrxtypac

Exercise: Display recent Ruby events

You can get the list of events from https://api.github.com/repos/ruby/ruby/events

If you learn enough jq, this is where it shines, here's my jq solution:

$ curl -sL https://api.github.com/repos/ruby/ruby/events \
>  | jq -cr 'map([.created_at, .type, .actor.display_login]|join(" "))[]' \
>  | column -t
2017-05-28T16:29:53Z  PushEvent                      hsbt
2017-05-28T14:18:16Z  PushEvent                      hsbt
2017-05-28T13:27:47Z  CommitCommentEvent             MSP-Greg
...

But I've found it rather difficult to remember how it works right when I need it, and I do know Ruby without having to go look anything up, so here's my Ruby solution:

$ curl -sL 'https://api.github.com/repos/ruby/ruby/events' \
>  | ruby -r json -e 'JSON.parse($stdin.read).map { |e| puts "#{e["created_at"]} #{e["type"]} #{e["actor"]["display_login"]}" }' \
>  | column -t
2017-05-28T16:29:53Z  PushEvent                      hsbt
2017-05-28T14:18:16Z  PushEvent                      hsbt
2017-05-28T13:27:47Z  CommitCommentEvent             MSP-Greg
...