[ https://issues.apache.org/jira/browse/ARROW-11994?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17302979#comment-17302979 ]
Ian Cook commented on ARROW-11994: ---------------------------------- There was a brief discussion of this dependency of Dataset on Parquet at https://issues.apache.org/jira/browse/ARROW-11735?focusedCommentId=17291960&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17291960 but the outcome of that was simply that we allowed both Dataset and Parquet to be toggled off in the R package build. (The original scope of ARROW-11735 was solely Dataset.) One simple solution would be to detect that Dataset is enabled and Parquet is not and fail the build with a helpful message indicating you must enable both or neither. > [R] Build fails if dataset enabled but parquet is not > ----------------------------------------------------- > > Key: ARROW-11994 > URL: https://issues.apache.org/jira/browse/ARROW-11994 > Project: Apache Arrow > Issue Type: Bug > Components: R > Reporter: Neal Richardson > Priority: Minor > > Following ARROW-11735; discovered while working on ARROW-10734. The > arrow::dataset::ParquetFileFormat and related classes require both dataset > and parquet. The {{#if defined}} logic in r/src/dataset.cpp is right and both > are required, but in the wrapping that is generated for arrowExports.cpp, we > only use the annotation on the functions, {{[[dataset::export]]}} to wrap. So > the ParquetFileFormat methods in arrowExports.cpp are if defined > ARROW_R_WITH_DATASET and fail if parquet is not available. > Not a priority to fix (for Solaris I can turn off ARROW_DATASET and avoid > this), just wanted to note it in case we need to revisit this wrapping logic > later anyway. cc [~icook] -- This message was sent by Atlassian Jira (v8.3.4#803005)