null values

[Jarosław Pękala <jaroslaw.pekala@…>]

When data contains nulls like:

Sp_data	Ideal	Real
2013-06-22 00:00:00+02:00	39	39
2013-06-22 00:00:00+02:00		39
2013-06-23 00:00:00+02:00	36
2013-06-24 00:00:00+02:00	33	32

error is raised:

All series on a given axis must be of the same data type

when null value is replaced by exact value ex:

Sp_data	Ideal	Real
2013-06-22 00:00:00+02:00	39	39
2013-06-22 00:00:00+02:00	1	39
2013-06-23 00:00:00+02:00	36	1
2013-06-24 00:00:00+02:00	33	32

chart is ok

I added to [vis] in trac.ini:

options = interpolateNulls:true

but result is the same as without this line.

[Jarosław Pękala <jaroslaw.pekala@…>]

Two causes:

1. Column types is deduced from first row values. If some of numeric columns contain 'null', column is treated as string. If mix of series types (numeric & string) is sent to API, it will result in an All series on a given axis must be of the same data type error. Workaround: first row cannot have null in non-string column.
2. Plugin does not correctly support null values. Consider example:

Col1	Col2	Col3
'2014-01-01'	10	10
'2014-01-02'	null	9

function populate (viz.js) will initialize DataTable with Col1 (date), Col2 (numeric), Col3 (numeric). First row will be added properly. Second will be skipped because Col2 = null, which is deduced as string - see line:

if (isNaN(value)) {
   // not a number, assume a string
   value = raw;
   type = 'string';
} 

because of line:

if (type != data.getColumnType(j))
    goodrow = false;

row is marked as invalid

I used workaround by adding one condition:

// if first row, add the columns
            if (i==0)
                data.addColumn(type, columns[j]);
	    else if (raw == '')
	    {
		value = null;
		goodrow = true;
	    }
            else if (type != data.getColumnType(j))
                goodrow = false;

This is not the solution for the point 1 (types) but solves the second (null)