Javier Luraschi created SPARK-25902:
---------------------------------------
Summary: Support for dates with milliseconds in Arrow bindings
Key: SPARK-25902
URL: https://issues.apache.org/jira/browse/SPARK-25902
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 2.3.2
Reporter: Javier Luraschi
Currently, the Apache Arrow bindings for Java only support `Date` with the
metric set to `DateUnit.DAY`, see
[ArrowUtils.scala#L72|https://github.com/apache/spark/blob/8c2edf46d0f89e5ec54968218d89f30a3f8190bc/sql/core/src/main/scala/org/apache/spark/sql/execution/arrow/ArrowUtils.scala#L72].
However, the Spark Arrow bindings for `R` are adding support to map `POSIXct`
to `Date` using `DateUnit.MILLISECOND`, with the following code triggering the
following warning:
{code:java}
devtools::install_github("apache/arrow", subdir = "r")
devtools::install_github("rstudio/sparklyr", ref = "feature/arrow")
Sys.setenv("SPARK_HOME_VERSION" = "2.3.2")
library(sparklyr)
library(arrow)
sc <- spark_connect(master = "local", spark_home = "<path-to-spark-sources>")
dates <- data.frame(dates = c(
as.POSIXlt(Sys.time(), "GMT"),
as.POSIXlt(Sys.time(), "EST"))
)
dates_tbl <- sdf_copy_to(sc, dates, overwrite = T){code}
{code:java}
Arrow disabled due to columns: dates
{code}
Which means that Arrow serialization gets disabled due to the following Spark
exception being thrown:
{code:java}
java.lang.UnsupportedOperationException: Unsupported data type:
Date(MILLISECOND)
{code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
