spark-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Hyukjin Kwon <gurwls...@gmail.com>
Subject Re: Support for decimal separator (comma or period) in spark 2.1
Date Thu, 23 Feb 2017 12:56:04 GMT
Please take a look at https://issues.apache.org/jira/browse/SPARK-18359.

2017-02-23 21:53 GMT+09:00 Arkadiusz Bicz <arkadiusz.bicz@gmail.com>:

> Thank you Sam for answer, I have solved problem by loading all decimals
> columns as string and replacing all commas with dots but this solution is
> lacking of automatic infer schema which is quite nice functionality.
>
> I can work on adding new option to DataFrameReader  for localization like:
>
> spark.read.option("NumberLocale", "German").csv("filefromeurope.csv")
>
> Just wonder if it will be accepted?
>
> Best Regards,
>
> Arkadiusz Bicz
>
> On Thu, Feb 23, 2017 at 1:35 PM, Sam Elamin <hussam.elamin@gmail.com>
> wrote:
>
>> Hi Arkadiuz
>>
>> Not sure if there is a localisation ability but I'm sure other will
>> correct me if I'm wrong
>>
>>
>> What you could do is write a udf function that replaces the commas with a
>> .
>>
>> Assuming you know the column in question
>>
>>
>> Regards
>> Sam
>> On Thu, 23 Feb 2017 at 12:31, Arkadiusz Bicz <arkadiusz.bicz@gmail.com>
>> wrote:
>>
>>> Hi Team,
>>>
>>> I would like to know if it is possible to specify decimal localization
>>> for DataFrameReader for  csv?
>>>
>>> I have cvs files from localization where decimal separator is comma like
>>> 0,32 instead of US way like 0.32
>>>
>>> Is it a way to specify in current version of spark to provide
>>> localization:
>>>
>>> spark.read.option("sep",";").option("header",
>>> "true").option("inferSchema", "true").format("csv").load("no
>>> nuslocalized.csv")
>>>
>>> If not should I create ticket in jira for this ? I can work on solution
>>> if not available.
>>>
>>> Best Regards,
>>>
>>> Arkadiusz Bicz
>>>
>>>
>>>
>

Mime
View raw message