Several suggestions for improving tabular data visualisation/exploration

[sylvaticus]

This is a set of suggestions and notes of things that still do not match Atom in the topic of (tabular) data exploration.

To show my points I am going to use the following script:

using TableView, DataFrames

sm      = [1 2 3; 4 5 6]
bigm    = rand(1000,1000)
wrapper = [[sm,sm,sm] [bigm,bigm,bigm]]
mixm    = [[1,2,3] ['a','b','c']]
sdf     = DataFrame(a = [1,2,3], b=['a','b','c'])
bigdf   = DataFrame(bigm)
showtable(sm)
showtable(bigm)
showtable(wrapper)
showtable(mixm)
showtable(sdf)
showtable(bigdf)

This is how it looks like in atom:

We note that that the Workspace does a good job in showing the data, maintaining its shape. However the main drawback is that the tables are displayed (with the external package TableView) in the unique "Plot" panel, meaning we can visualise only one table at the time.

This is how it appears to work on the same script in VS Code:

We see that the the data in the Workspace is showed with a drill-out approach where the whole table and shape is instead rendered using a specific icon that open the data on a different panel. This is great as we can visualise several tables side-by-side !
There are however a few problems that I would like to highlight:

Tables titles are fixed to "Julia Table"
It would be much better to have as title of each panel the specific variable name

Displayed tables/df don't seem to have a minimum size and don't have row headers
In contrast to Atom, the matrix or df displayed don't seem to have a minimum width/height, so they are useless for big tables.
Also, both matrices and dataframes don't have row numbers like instead in Atom/TableViews.

Df tables have column name capitalised
As Julia (like almost all programming languages) is case sensitive, I think it is best to keep this info in the visualised df header.

Workspace panel is over-bloated
While one can agree to have a drill-out approach as the tabular visualisation is integrated in VS code, I think the label "Cartesian Index" could be suppressed. Maybe its space could be used to display the two actual values at the two extremes of the range, e.g. [21,8,1]...[40,8,1] 0.621252...0.241511

Array{Any,2} are not displayed
[EDIT]: This happens only if done with "using TableView", so it's not actually a problem.
(see the warning in the console). This seems to be platform dependant. I am using Julia for VS Code 1.0.2 with Julia 1.4.1 in Ubuntu Linux.

Tabular display doesn't support drilling
This is very nice in R-Studio/Spyder: if the content of a cell is an other collection (array, matrix) clicking on it would expand and display it recursively. This is not currently supported neither in VS Code nor in Atom

Big tables take a while to open
In the script this is true for bigm but also for wrapper that has only 6 cells. Some sort of threashold-size based lazy loading would help

Thank you for the development of VS Code. Having the possibility to see different tables in one windows is great for data exploration!
This give me idea for an other request, not really on tabular but still on data exploration:

Allow decoupling of plot panels
Sometime you want to see plots side to side. It would be nice to have a button "Open in a new panel" on plot panels.

Thank you again.