Support pattern matching in CSV rows

[kddnewton]

No description provided.

[baweaver]

The keys, as supplied, will be Symbols. This can present an issue as CSV rows will more than likely be String keyed:

require 'csv'
# => true

csv = CSV.parse <<~RUBY, headers: true
  a,b,c
  1,2,3
  2,3,4
  4,5,6
RUBY
# => #<CSV::Table mode:col_or_row row_count:4>

csv.headers
# => ["a", "b", "c"]

csv.to_a
# => [["a", "b", "c"], ["1", "2", "3"], ["2", "3", "4"], ["4", "5", "6"]]

csv.first.class
# => CSV::Row

csv.first.fields
# => ["1", "2", "3"]
csv.first.headers
# => ["a", "b", "c"]

If you want to support this you would want to instead transform the keys to String, but this may be controversial as it conflates the two.

[baweaver]

This could be resolved one of a few ways.

In the first, retrieval of values would be done versus the String equivalent of the value:

def deconstruct_keys(keys)
  keys.to_h { |k| [k, self[k.to_s]] }
end

This assumes, however commonly it may be, that all keys are String values. This could be resolved via:

def deconstruct_keys(keys)
  first_header = headers.first
  case first_header
  when String then keys.to_h { |k| [k, self[k.to_s]] }
  when Symbol then keys.to_h { |k| [k, self[k]] }
  else {}
  end
end

...though this is far more verbose. I do not think that Symbol keys are commonly used in the context of CSV, so this concern might be ignored safely, but I leave that up to the core group of CSV to decide the risk.

[kddnewton]

I think I'd rather leave it as is for now. You can always pass header_converters: :symbol to the parse function, which would make this work as expected.

I think I would be really surprised if in my matching I specified symbols and it matched against strings.

[kou]

Sorry for not responding this.
Could you add support for in **rest case?

[kou]

Could you add one more test case for in B:, **rest?

[kddnewton]

Updated!

[kou]

Thanks!