[BUG] labelOpts undefined

[Pimnr47]

on line 11546 in dist of 2.7.1:

var fontSize = valueOrDefault(labelOpts.fontSize, globalDefault.defaultFontSize);

labelOpts is undefined.

I just followed the basic getting started:

http://www.chartjs.org/docs/latest/getting-started/usage.html

[etimberg]

@Pimnr47 do you have a fiddle that reproduces this as well? Might help in debugging

[etimberg]

@Pimnr47 I created a fiddle with the same code as the page linked above and no errors are seen in the console.

[jcopperfield]

I think @Pimnr47 has some code that sets the legend.labels option to something undefined. Maybe as the result of some function.

Example to trigger issue

var ctx = document.getElementById("myChart");
var legendOptions = {};
var myChart = new Chart(ctx, {
    type: 'bar',
    data: {
        labels: ["One", "Two", "Three", "Four", "Five", "Six"],
        datasets: [{
            label: '# of Votes',
            data: [12, 19, 3, 5, 2, 3],
        }]
    },
    options: {
        legend: {
            labels: legendOptions.labels
        },
    }
});

To prevent any such cases being reported in the future I think the configuration should be sanitized before being merged with the defaults, or warn the user with console.warn.

[etimberg]

Thanks @jcopperfield! I think we could put a console.warn statement just before https://github.com/chartjs/Chart.js/blob/master/src/helpers/helpers.core.js#L189 which should produce a warning statement if sval is undefined.

@simonbrunel @benmccann thoughts?

[jcopperfield]

@etimberg I'm not sure that will totally solve the problem. For such a checker to be robust, it should check every user supplied setting being the correct type and respecting its limits.
e.g. if someone sets the barPercentage percentage it should be checked to be a Number and that its value is in the range [0.0, 1.0]. If it's not, use the default and maybe possibly use console.warn to inform the user/developer. This checking should of course be repeated before every update, possibly also checking the datasets for invalid values to prevent errors like #4978.

[simonbrunel]

@etimberg IIRC, sval undefined is a valid input for the merge process (e.g. to "invalidate" an option), which seems what happen here. labelOpts === undefined is not supported, but merge({a: 42}, {a: undefined}) looks correct to me, so I would not change the merge logic.

@jcopperfield Actually, barPercentage may become a scriptable options, so would support Number|Number[]|Function, the function returning Number|Number[] depending on the current data, which complicate the validation process.

If we decide to validate user inputs, we should do that for every single option (for consistency), not only for reported issues. Though, that could be complicated, for example with scriptable options where we need to evaluate a function, in a specific data/dataset context, to get the final value. I feel that will make the code much slower but also chaotic if we don't have a centralized/declarative way to validate options.

[jcopperfield]

@simonbrunel I agree that having Function as scriptable option would complicate a centralized sanitizing function. In that case the validation should be done at runtime, since the function has to be called at some point, maybe by wrapping it in a helper-function.

[simonbrunel]

Doing such validation for every options at every data (scriptable options) will seriously degrade update performance, I'm not sure it worths it :\

[jcopperfield]

Just checking the type wouldn't be much slower than the helper-function valueOrDefault, which is already used. I agree that checking scriptable options is less important, although they are more error prone, so some sanitation has to be done before use anyway. Or maybe we are talking about two different things.

[sweatherall]

I'm running into the same issue, after attempting to follow the basic example.
This is the instantiation code I have:

var myChart = new Chart(metric_chart, {
                        type: 'line',
                        // The data for our dataset
                        data: {
                            labels: ["January", "February", "March", "April", "May", "June", "July"],
                            datasets: [{
                                label: "My First dataset",
                                backgroundColor: 'rgb(255, 99, 132)',
                                borderColor: 'rgb(255, 99, 132)',
                                data: [0, 10, 5, 2, 20, 30, 45],
                            }]
                        },
                        options: {}
                    });

[jcopperfield]

@sweatherall Your example seems to be working for me (example)