Re: [bug-gawk] Tentative CSV extension - please advise

[Hermann Peifer]

On 2016-03-15 15:54, Miriam English wrote:
> 
> address@hidden wrote:
> 
>> Yes. That's getting too messy.  Instead such an extension should
>> simply define
>> a csvsplit() function.
>>
>> The big problem is embedded newlines in the fields.  Sigh. If not for
>> that
>> we could deal with this issue much more easily...
> 
> If the fields are preprocessed by turning
> "the quick
> brown fox"
> into
> "the quick\nbrown fox"
> then it becomes much easier. Then after fields are extracted the escaped
> characters can be returned to their literals.
> 
> It would be slightly simplified using something similar to what sed
> recently has done in adding the -z option, where the input "line" is
> delimited by a zero byte instead of a newline. I often use that to
> process entire files as a single line, treating newlines as just another
> character.
> 
> It reduces the problem to working out which newlines end a csv record
> and shouldn't be escaped.
> 
> 

I assume people know http://lorance.freeshell.org/csv/csv.txt, which
works fine, as far as I can tell from occasional usage.

As big problem, I rather remember quotes inside quotes.

Hermann