Best way to plot GraphBuilder plot with multiple Y variables using JSL

[#1]

All, 

I'd like to ask for an input on the best way to script a GB plot with multiple parameters.

Let's say I have a table like this:

New Table( "Untitled 140",
	Add Rows( 12 ),
	New Column( "Category",
		Character,
		"Nominal",
		Set Values( {"A", "A", "A", "B", "B", "B", "C", "C", "C", "D", "D", "D"} )
	),
	New Column( "Value1",
		Numeric,
		"Continuous",
		Format( "Best", 12 ),
		Set Values( [1, 4, 2, 3, 6, 2, 3, 5, 4, 1, 3, 6] )
	),
	New Column( "Value2",
		Numeric,
		"Continuous",
		Format( "Best", 12 ),
		Set Values( [6, 3, 5, 2, 8, 6, 5, 2, 6, 3, 5, 4] )
	)
)

And I want to script a plot like this:

where the autogenerated script looks like this:

Graph Builder(
	Size( 522, 450 ),
	Show Control Panel( 0 ),
	Variables( X( :Category ), Y( :Value1 ), Y( :Value2, Position( 1 ) ) ),
	Elements( Bar( X, Y( 1 ), Y( 2 ), Legend( 4 ), Summary Statistic( "Range" ) ) )
);

with the catch that there is an arbitrary number of Y columns and the names are also arbitrary.

For plots like Variability plot I would something like this:

parameters = {:"Value1"n, :"Value2"n};
yExpr = Expr( Y() );
For Each( {value, index}, parameters, Insert Into( yExpr, Name Expr( value ), 1 ) );

Eval(
	Substitute(
			Expr(
				Variability Chart( placeholder, X( :Category ),  Std Dev Chart( 0 )  )
			),
		Expr( placeholder ), Name Expr( yExpr )
	)
);

where I receive the names of the columns as a list and then construct a Y expression and then put it into the script for the chart. It would give me a list of variability plots, but that's OK with me.

If I want to take the same route in case of the GB it looks like I will have to construct two expressions:

Y( :Value1 ), Y( :Value2, Position( 1 ) )

and 

Elements( Bar( X, Y( 1 ), Y( 2 ), Legend( 4 ), Summary Statistic( "Range" ) ) )

Which is essentially the whole script.

My question: Is there a better way rather than making bunch of substitutions with expressions to assemble those two lines?

Thanks!

[#2]

Ok, this is how I solved it, but is there a simpler way? Can't believe that to script such a simple thing I need to go into that kind of complexity...:

parameters = {:"Value1"n, :"Value2"n};

yExpr = Expr( Y() );
yExprPos = Expr( Y( Position( 1 ) ) );
barExpr = Expr(
	Bar( X, Summary Statistic( "Range" ) )
);
yNExpr = Expr( Y() );
variablesExpr = Expr(
	Variables( X( :Category ) )
);

For Each( {value, index}, parameters,
	Insert Into( yExpr, Name Expr( value ), 1 );
	yExprPos = Expr( Y( Position( 1 ) ) );
	Insert Into( yExprPos, Name Expr( value ), 1 );
	Insert Into( variablesExpr, Name Expr( yExprPos ) );
	//This is for Y(1) etc. in Bar()
	yNExpr = Expr( Y() );
	Insert Into( yNExpr, Name Expr( index ) );
	
	Insert Into( barExpr, Name Expr( yNExpr ), 2 );
);


//Name Expr( yExpr );
Show(Name Expr( variablesExpr ));
Show(Name Expr( barExpr ));

gbExpr = Expr(
	Graph Builder( Size( 522, 450 ), Show Control Panel( 0 ), placeholder1, Elements( placeholder2 ) )
	
);

Eval( 
	Substitute( 
		Substitute( 
			Name Expr(gbExpr),    
			Expr( placeholder1 ), 
			Name Expr( variablesExpr )
		), 
		Expr( placeholder2 ), 
		Name Expr( barExpr ) 
	) 
);

[#3]

I have usually used similar methods as you to build such graph builders, this might be a good readRe: Can you construct this without writing expression a string? 

[#4]

This still uses loops to build those expressions, but it concatenates lists of arguments instead of substituting them:

Names default to here(1);

//call another function (f) with the arguments provided in a list (l)
do.call = function({l,f}, 
	Substitute(
		Name Expr(l),
		Expr({}),
		Name Expr(f)
	);
);

dt = New Table( "Data",
	Add Rows( 12 ),
	New Column( "Category", Character, "Nominal",
		Set Values( {"A", "A", "A", "B", "B", "B", "C", "C", "C", "D", "D", "D"} )
	),
	New Column( "Value1", Numeric, "Continuous", Format( "Best", 12 ),
		Set Values( [1, 4, 2, 3, 6, 2, 3, 5, 4, 1, 3, 6] )
	),
	New Column( "Value2", Numeric, "Continuous", Format( "Best", 12 ),
		Set Values( [6, 3, 5, 2, 8, 6, 5, 2, 6, 3, 5, 4] )
	)
);

vars = {:Value1, :Value2};

VarExpr = do.call(
	{ X( :Category ) } || transform each({v}, vars, eval expr( Y( Expr( Name Expr( v ) ), Position( 1 ) ) ) ),
	expr(Variables())
);

BarExpr = do.call(
	{ X } || transform each({v, i}, vars, eval expr( Y( Expr( i ) ) ) ) || { Legend( 4 ), Summary Statistic( "Range" ) },
	expr(Bar())
);

Eval( Eval Expr(
	dt << Graph Builder(
		Size( 522, 450 ),
		Show Control Panel( 0 ),
		Expr( Name Expr( VarExpr ) ),
		Elements( Expr( Name Expr( BarExpr ) ) )
	)
) );

[#5]

Don't forget the STACK command.  Requires eval(parse()) but it's pretty clean otherwise.  Here's an example with three columns to be plotted.

dt = New Table( "Test",
	Add Rows( 12 ),
	New Column( "Category",
		Character,
		"Nominal",
		Set Values( {"A", "A", "A", "B", "B", "B", "C", "C", "C", "D", "D", "D"} )
	),
	New Column( "Value1",
		Numeric,
		"Continuous",
		Format( "Best", 12 ),
		Set Values( [1, 4, 2, 3, 6, 2, 3, 5, 4, 1, 3, 6] )
	),
	New Column( "Value2",
		Numeric,
		"Continuous",
		Format( "Best", 12 ),
		Set Values( [6, 3, 5, 2, 8, 6, 5, 2, 6, 3, 5, 4] )
	),

	New Column( "Value3",
		Numeric,
		"Continuous",
		Format( "Best", 12 ),
		Set Values( [1, 4, 3, 6, 5, 7, 3, 4, 2, 8, 6, 1] )
	)
);

dt_cols = dt << get column names(string);
removefrom(dt_cols, 1);	// Remove Category

// Create the list of columns, like this: ":Value1, :Value2, :Value3"
col_list = ":" || concat items(dt_cols, ", :");

stack_str = evalinsert(
"\[dts = dt << Stack(invisible,
	columns( ^col_list^ ),
	Source Label Column( "Label" ),
	Stacked Data Column( "Data" ))]\");
eval(parse(stack_str));

dts << Graph Builder(
	Size( 518, 448 ),
	Show Control Panel( 0 ),
	Variables( X( :Category ), Y( :Data ), Overlay( :Label ) ),
	Elements( Bar( X, Y, Legend( 7 ) ) )
);

[#6]

@pmroz here is the same code without Eval(Parse()).  I agree this is much cleaner!

Names default to here(1);

dt = New Table( "Test",
	Add Rows( 12 ),
	New Column( "Category", Character, "Nominal",
		Set Values( {"A", "A", "A", "B", "B", "B", "C", "C", "C", "D", "D", "D"} ) ),
	New Column( "Value1", Numeric, "Continuous", Format( "Best", 12 ),
		Set Values( [1, 4, 2, 3, 6, 2, 3, 5, 4, 1, 3, 6] ) ),
	New Column( "Value2", Numeric, "Continuous", Format( "Best", 12 ),
		Set Values( [6, 3, 5, 2, 8, 6, 5, 2, 6, 3, 5, 4] ) ),
	New Column( "Value3", Numeric, "Continuous", Format( "Best", 12 ),
		Set Values( [1, 4, 3, 6, 5, 7, 3, 4, 2, 8, 6, 1] ) )
);

dt_cols = dt << get column references();
remove from(dt_cols, 1);	// Remove Category

dts = dt << Stack(invisible,
	columns( dt_cols ),
	Source Label Column( "Label" ),
	Stacked Data Column( "Data" )
);
	
dts << Graph Builder(
	Size( 518, 448 ),
	Show Control Panel( 0 ),
	Variables( X( :Category ), Y( :Data ), Overlay( :Label ) ),
	Elements( Bar( X, Y, Legend( 7 ) ) )
);

[#7]

Nice!  Thanks for adding that touch.  Wish I'd have tried that myself!