Re: CSV extension status

[Manuel Collado]

El 16/05/2021 a las 22:21, Ed Morton escribió:
> The gawkextlib CSV extension has been listed on
> http://gawkextlib.sourceforge.net/ for a while now (a year or more?) but
> noted as "(not yet released)". Is there anything specific holding up its
> release and/or a planned date when it will be released?

Thanks for your interest.

The gawk-csv extension pushes the gawk API for input parsers to its 
limits. And uncover some nasty limitations. An input parser can deliver 
a record composed of fields, but there is no way to control further 
reparsing of the record after assigning new values to $0..$NF. To do so 
it it necessary to have a companion pure (g)awk library that temporarily 
overrides FS, OFS, etc. In addition the generation of CSV output is 
easier to code in AWK than in C.

The fact is that the combination of FPAT and BEGINFILE/ENDFILE has 
enough power to implement an effective CSV processing with just a pure 
gawk library. Please look at the CSVMODE library available at:

    http://mcollado.z15.es/xgawk/

This library extends the functionality of the gawk-csv extension with 
the ability to operate with either CSV fragments or clean text field values.

The only advantage of using the API for parsing CSV fields is the 
ability to provide meaningful error messages and recovering of malformed 
CSV data.

In order to made a decision about maintenance of the gawk-csv extension 
a feedback from potential users will be very welcome. For instance:

- Should it be desirable to also host pure gawk modules in the gawkextib 
site?

- In order to unify the user interface of gawk-csv and the CSVMODE 
library, which control variable value is more natural to select clean 
text field values?
   CSVMODE = 1 as in gawk-csv, or
   CSVMODE = -1 as in the CSVMODE library ?

Any opinion or suggestions about this or similar extensions will be very 
appreciated.

Regards.

--
Manuel Collado - http://mcollado.z15.es