Toby Dylan Hocking

no need for hash table. because the data are so small, there is no need for efficiency. I suggest using a data table ```r > tds=rbind(data.table(row=1,col=1,rowspan=1,colspan=1), data.table(row=0,col=0,rowspan=2,colspan=2)) > (cells=tds[, CJ(row=row+seq(0,rowspan-1),...

I think this is ok, because your example is not a typical use case (we don't care about how the renderer works in that case because we don't expect to...

wow dynamic programming I don't think such advanced functionality is necessary though we don't need to support non-rectangular layouts, with one plot outside the main rectangle. again the use case...

this is getting too complicated. I think we should simplify the interface to ease maintenance of the code. I propose that the user should not be able to specify `col`...

what about only allowing user to define rowspan and colspan for each plot (not row nor col), and then we define "tr" in animint? ```r animint( "tr", ggplot()+... "tr", ggplot()+......

Maybe I don't understand the issue you described as "even if we remove the option of row and column and just keep rowspan and colspan, we will still face issue...

hi @siddhesh195 I think your construct_outer_table function is un-necessarily complex. please see an alternative in #153 which is much simpler. If you agree, please close this PR and push updates...

superseded by #153

thanks for the update. Is it normal that there are so many files changed? I see a lot of image files updated, but the images do not look any different...

even if the total file size is the same before and after the PR, the total git repo size doubles ...