Converting ruby hashes to JSON that is both valid and human readable - ruby

Question

I have code that works but am soliciting suggestions for improvement.
I have a file containing ruby hashes:
{"dat"=>"2013-09-01T20:40:00-07:00", "sca"=>"5", "del"=>"755", "dir"=>"S"}
{"dat"=>"2013-09-01T21:00:00-07:00", "sca"=>"5", "del"=>"459", "dir"=>"S"}
that I want to convert to JSON that is both valid and human-readable. This code is compact and produces valid JSON...
#!/usr/bin/env ruby
# expected input: file of hashes, one/line
# output: properly formatted json array
require 'json'
json_array = []
while input = ARGF.gets
input.each_line do |line|
json_array.push( eval(line) )
end
end
print json_array
puts
..but without any newlines is not easily human-readable:
[{"dat"=>"2013-09-01T20:40:00-07:00", "sca"=>"5", "del"=>"755", "dir"=>"S"}, {"dat"=>"2013-09-01T21:00:00-07:00", "sca"=>"5", "del"=>"459", "dir"=>"S"}]
Substituting
puts JSON.pretty_generate(json_array)
for the two output lines above produces valid JSON that is human-readable, but verbose:
[
{
"dat": "2013-09-01T20:40:00-07:00",
"sca": "5",
"del": "755",
"dir": "S"
},
(more lines...)
Better from a human-readbiility standpoint would be to have a "record" on each line:
[
{"dat":"2013-09-01T20:40:00-07:00","sca":"5","del":"755","dir":"S"},
{"dat":"2013-09-01T21:00:00-07:00","sca":"5","del":"459","dir":"S"}
]
But in order to avoid the trailing comma issue [apparently a common problem - see http://trailingcomma.com/ ] I have resorted to an ugly loop with special casing. While it accomplishes the goal, I'm not happy about it and I feel like there must be a simpler way:
#!/usr/bin/env ruby
# expected input: file of hashes, one/line
# output: properly formatted json array
require 'json'
prevHash = ""
currHash = ""
puts "["
while input = ARGF.gets
# in order to to prevent a dangling comma on last element in output json array
# this counter-intuitive loop always outputs the prev, not the current, array elem
# with a trailing comma
input.each_line do |currLine|
currHash = eval(currLine) # convert string to hash
if (prevHash != "") # if not first time thru
puts " " + prevHash.to_json + ","
end
prevHash = currHash
end
end
# then, finally add the last array element *without* the troublesome trailing comma
puts " " + currHash.to_json
puts "]"
Suggestions welcome, particularly those that show me the artful one-liner that I missed.

Answer 1

JSON.pretty_generate accepts an optional hash parameter where you can configure the generator.
A state hash can have the following keys:
indent: a string used to indent levels (default: ”),
space: a string that is put after, a : or , delimiter (default: ”),
space_before: a string that is put before a : pair delimiter (default: ”),
object_nl: a string that is put at the end of a JSON object (default: ”),
array_nl: a string that is put at the end of a JSON array (default: ”),
allow_nan: true if NaN, Infinity, and -Infinity should be generated, otherwise an exception is thrown if these values are encountered. This options defaults to false.
max_nesting: The maximum depth of nesting allowed in the data structures from which JSON is to be generated. Disable depth checking with :max_nesting => false, it defaults to 19.
Playing around with that the closest I could get to your requirement is
JSON.pretty_generate(hash, {object_nl: '', indent: ' '})
which renders to
[
{ "dat": "2013-09-01T20:40:00-07:00", "sca": "5", "del": "755", "dir": "S"},
{ "dat": "2013-09-01T21:00:00-07:00", "sca": "5", "del": "459", "dir": "S"}
]