Possible to use where clause with regex or string search?

[baarkerlounger]

df = Daru::DataFrame.new({y1: ["action|thriller", "comedy", "Drama"], y2: [9, 10, 3]}, 
     index: ["a", "b", "c"])

Looking to do something along the lines of df.where(df[:y1].contains('action'))

Pandas has df['y1'].str.contains('action')

My current workaround looks like:

Daru::Core::Query::BoolArray.new(df[:y1].map{ |e| e.to_s.include?("action") })

[zverok]

You can use more Ruby-idiomatic filter(:row):

df.filter(:row) { |r| r[:y1].include?('action') }
# => #<Daru::DataFrame(1x2)>
#                    y1         y2
#          a action|thr          9 

This idea of bool arrays was borrowed from pandas, but it would be probably retired in future, in favor of more idiomatic ways.

(Though, I should say that filter currently is slower)

[baarkerlounger]

The bool array has the advantage of being easy to &/| with other filtering/slicing methods. Neither version seems particularly nice compared to the Pandas way though.

[zverok]

Hm. Can you please show how pandas is more powerful? For me it looks like natural Ruby's blocks can do everything (just can be pretty slow on it)

[baarkerlounger]

I'm not saying more powerful necessarily but I think nicer to work with (and performance). Compare:

Pandas: filtered = df[(df['y1'] == 'Movie') & (df['y2'].str.contains('Action'))] Daru: filtered = df.where(df[:y1].eq('Movie')).filter(:row) { |r| r[:y2].include?('Action')}

The first example to me it's more readable my filter has 2 conditions.

[zverok]

filtered = df.filter(:row) { |r| r[:y1] == 'Movie' && r[:y2].include?('Action')}

...just like you'd filter your usual arrays.