[ 
https://issues.apache.org/jira/browse/CSV-110?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13960973#comment-13960973
 ] 

Gabriel Reid commented on CSV-110:
----------------------------------

Thanks for the info [~garydgregory].

The main use case that I'm interesting in supporting (and which I didn't make 
very clear in the description) is the concept of being able to reuse a single 
CSVLineParser to parse multiple lines of input which aren't available as as 
single contiguous stream.

As you pointed out, it's currently easy to parse a single line, but this single 
line parsing also requires instantiating a new CSVParser for every input line. 
Although it may seem like a micro-optimization in trying to get around this 
CSVParser instantiation for each input line, I believe it's worth the effort 
when parsing many billions of input records (which is a common use case when 
working with Hadoop).

> Add ability to parse single lines
> ---------------------------------
>
>                 Key: CSV-110
>                 URL: https://issues.apache.org/jira/browse/CSV-110
>             Project: Commons CSV
>          Issue Type: New Feature
>            Reporter: Gabriel Reid
>         Attachments: CSV-110.patch
>
>
> Due to the iterator-based API of CSVParser, there is currently no simple and 
> convenient way to parse single lines of CSV-formatted data. The intention of 
> this ticket is to add something along the lines of the following:
> {code}
> CSVLineParser lineParser = new CSVLineParser(csvFormat);
> String singleLine = "a,b,c";
> CSVRecord singleRecord lineParser.parseLine(singleLine);
> {code}
> The use case of parsing single lines comes up very often in terms of 
> distributed batch processing scenarios (i.e. Hadoop jobs), and CSV-style 
> formats are also regularly used in such scenarios. Currently, projects are 
> often forced to build their own ad-hoc CSV parsing solutions, so adding the 
> ability to parse single lines to commons-csv would be very useful to these 
> projects, as well as anyone doing parsing based on input that isn't necessary 
> in the form of a single stream.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Reply via email to