#convert_fields's sym/string break isn't clarified in #initialize's docs about csv HOT 5 CLOSED

Can you show a runnable Ruby script that reproduces this case?

from csv.

@kou

raw_csv_data = "col_header_1, col_header_2\nval_1, val_2\n"

# This should raise and break execution flow if it is called
custom_converter = lambda { |header_val| raise 'break'}

broken_converters = [:symbol, custom_converter]
working_converters = [custom_converter, :symbol]

# custom_converter not called (as seen via lack of raise)
# symbol converter happens first, and short-circuits subsequent converters
CSV.parse(raw_csv_data, :headers => true, :header_converters => broken_converters)
# custom_converter called (as seen via raise)
CSV.parse(raw_csv_data, :headers => true, :header_converters => working_converters)

from csv.

Thanks.

In your use case (symbol -> squeeze("_") -> symbol), you should use :headers_converters =>[squeeze_converter, :symbol]. Because you can't call squeeze for Symbol. You can't put the squeeze converter after symbol converter.

:headers_converters => [:symbol, squeeze_converter] is a your application bug.

The current behavior is reasonable for this use case.
Do you have any other use case?

from csv.

@kou regardless of what my converter does, if its after the :symbol converted, it would never execute.

The docs for the CSV initializer method don't state this, and it seems unreasonable that a dev should have to dig way deep into the source code to discover that the header_converter loop early-exits after anything returns a symbol.

It seems like that early-exit is a bug, or, at the very least, needs to be better documented in the initializer docs.

from csv.

Can you try sending a pull request to improve the current document?

from csv.