[jira] [Updated] (ARROW-15833) [C++] Support casting from duration to string/utf8

["Dragoș Moldovan-Grünfeld (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/ARROW-15833?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Dragoș Moldovan-Grünfeld updated ARROW-15833:
---------------------------------------------
    Description: 
Cast from {{duration[s]}} to {{utf8}} is not supported. This surfaced in R as an error when trying to write to CSV.
{code:r}
arrow::write_csv_arrow(data.frame(time = as.difftime(1, units="secs")), "test.csv")
#> Error: NotImplemented: Unsupported cast from duration[s] to utf8 using function cast_string
{code}
 
{code:r}
library(arrow, warn.conflicts = FALSE)
#> See arrow_info() for available features
c <- Array$create(as.difftime(32, units = "secs"))
c
#> Array
#> <duration[s]>
#> [
#>   32
#> ]

c$cast(string())
#> Error: NotImplemented: Unsupported cast from duration[s] to utf8 using function cast_string
#> /Users/dragos/Documents/arrow/cpp/src/arrow/compute/function.cc:231  DispatchBest(&inputs)
{code}
I guess the desired output would be to cast to "current" units even if we lose the metadata. Casting to {{int64()}} works (in a similar, lossy way).
{code:r}
c$cast(int64())$cast(string())
#> Array
#> <string>
#> [
#>   "32"
#> ]

c$cast(duration(unit = "ms"))$cast(int64())$cast(string())
#> Array
#> <string>
#> [
#>   "32000"
#> ]
{code}

  was:
Cast from {{duration[s]}} to {{utf8}} is not supported. This surfaced in R as an error when trying to write to CSV.

{code:r}
arrow::write_csv_arrow(data.frame(time = as.difftime(1, units="secs")), "test.csv")
#> Error: NotImplemented: Unsupported cast from duration[s] to utf8 using function cast_string
{code}
 
{code:r}
library(arrow, warn.conflicts = FALSE)
#> See arrow_info() for available features
c <- Array$create(as.difftime(32, units = "secs"))
c
#> Array
#> <duration[s]>
#> [
#>   32
#> ]

c$cast(string())
#> Error: NotImplemented: Unsupported cast from duration[s] to utf8 using function cast_string
#> /Users/dragos/Documents/arrow/cpp/src/arrow/compute/function.cc:231  DispatchBest(&inputs)
{code}
I guess the desired output would be to cast to "current" units even if we lose the metadata
{code:r}
c$cast(int64())$cast(string())
#> Array
#> <string>
#> [
#>   "32"
#> ]

c$cast(duration(unit = "ms"))$cast(int64())$cast(string())
#> Array
#> <string>
#> [
#>   "32000"
#> ]
{code}


> [C++] Support casting from duration to string/utf8
> --------------------------------------------------
>
>                 Key: ARROW-15833
>                 URL: https://issues.apache.org/jira/browse/ARROW-15833
>             Project: Apache Arrow
>          Issue Type: Improvement
>          Components: C++, R
>    Affects Versions: 7.0.0
>            Reporter: Dragoș Moldovan-Grünfeld
>            Priority: Major
>
> Cast from {{duration[s]}} to {{utf8}} is not supported. This surfaced in R as an error when trying to write to CSV.
> {code:r}
> arrow::write_csv_arrow(data.frame(time = as.difftime(1, units="secs")), "test.csv")
> #> Error: NotImplemented: Unsupported cast from duration[s] to utf8 using function cast_string
> {code}
>  
> {code:r}
> library(arrow, warn.conflicts = FALSE)
> #> See arrow_info() for available features
> c <- Array$create(as.difftime(32, units = "secs"))
> c
> #> Array
> #> <duration[s]>
> #> [
> #>   32
> #> ]
> c$cast(string())
> #> Error: NotImplemented: Unsupported cast from duration[s] to utf8 using function cast_string
> #> /Users/dragos/Documents/arrow/cpp/src/arrow/compute/function.cc:231  DispatchBest(&inputs)
> {code}
> I guess the desired output would be to cast to "current" units even if we lose the metadata. Casting to {{int64()}} works (in a similar, lossy way).
> {code:r}
> c$cast(int64())$cast(string())
> #> Array
> #> <string>
> #> [
> #>   "32"
> #> ]
> c$cast(duration(unit = "ms"))$cast(int64())$cast(string())
> #> Array
> #> <string>
> #> [
> #>   "32000"
> #> ]
> {code}



--
This message was sent by Atlassian Jira
(v8.20.1#820001)