[ https://issues.apache.org/jira/browse/TOREE-380?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16048203#comment-16048203 ]
ASF GitHub Bot commented on TOREE-380: -------------------------------------- Github user rdblue commented on a diff in the pull request: https://github.com/apache/incubator-toree/pull/104#discussion_r121757204 --- Diff: scala-interpreter/build.sbt --- @@ -18,3 +18,4 @@ import sbt.Tests.{Group, SubProcess} */ libraryDependencies ++= Dependencies.sparkAll.value +libraryDependencies += "com.github.jupyter" % "jvm-repr" % "0.1.0" --- End diff -- I think we'll need to take a look after this PR at the other interpreters. PySpark, for example, should adhere to the common python convention to use _repr_html_ for objects that exist in the python process, and should use the Jupyter library for objects that are resident in the JVM (like DataFrames). The background on Jupyter's jvm-repr library is that I'm trying to work with the Jupyter community to standardize a system for libraries and kernels to collaborate on rich representations for the JVM. Vegas, for example, should register a function to display its objects so it doesn't have to be built into the kernel or bolted on by users. > Interpreters should be allowed to send results other than text/plain. > --------------------------------------------------------------------- > > Key: TOREE-380 > URL: https://issues.apache.org/jira/browse/TOREE-380 > Project: TOREE > Issue Type: Improvement > Reporter: Ryan Blue > Assignee: Ryan Blue > > Jupyter allows kernels to send results using different content types, but > this isn't allowed by Toree for its interpreters. This prevents custom > display logic. The basic problem is that {{ExecuteOutput}} is a {{String}} > and not a {{Map[String, String]}} like {{CellMagicOutput}}. -- This message was sent by Atlassian JIRA (v6.4.14#64029)