issue with RJSONIO about htmlwidgets HOT 6 CLOSED

As indicated in the answer on SO, this is NOT a bug in RJSONIO. It is just a difference in how jsonlite and RJSONIO treat data frames. We will be switching to jsonlite in the long run, as shiny adopts jsonlite, but until then it will be RJSONIO. You can use the htmlwidgets:::to_json function to convert your data frame to a list of records and then feed it to RJSONIO::toJSON to get the desired output.

from htmlwidgets.

We can't convert to jsonlite until Shiny converts to jsonlite, and Shiny
can't do this until it has all the right knobs to allow it to emulate some
of the behaviors of RJSONIO. We'd certainly all like to do this but it's
not really possible right now so I'd suggest you figure out a local
workaround.

On Fri, Feb 6, 2015 at 1:33 PM, Enzo [email protected] wrote:

I've just found a problem in the conversion(s) between data.table and
RJSONIO.
This is affecting the library I developed: rpivotTable here
https://github.com/smartinsightsfromdata/rpivotTable
I have documented the problem with reproducible code here
http://stackoverflow.com/questions/28372298/json-issue-bug-with-data-table-and-rjsonio-jsonlite
and reported it to the data.table project.
I think it strengthen the case for htmlwidgets conversion to adopting
jsonlite (which does not appear to have the issue).

—
Reply to this email directly or view it on GitHub
#77.

from htmlwidgets.

Solution is easy and should be in smartinsightsfromdata/rpivotTable#7 with extensive discussion/debate smartinsightsfromdata/rpivotTable#5.

from htmlwidgets.

The issue here is that @timelyportfolio pull is not a "solution" but is a bypass (of course no disrespect intended). The problem is occurring with iris dataset as iris has a dot in colnames (eliminating the dot eliminates the problem) and is at the edge between data.table, RJSONIO and htmlwidgets. I wonder how the other packages developed with htmlwidgets behave with data.table (or if all have the "forced" conversion to data.frame proposed by @timelyportfolio ).

from htmlwidgets.

I don't consider this a workaround until data.table has its own toJSON conversion function. Until then I think converting to a standard R data structure with data.table's own built-in function is a solution and not a hack. Nothing is lost since we lose data.table when we go to JavaScript anyways.

from htmlwidgets.

As an example just look at how much JJ has had to add to support dates in dygraphs.

from htmlwidgets.