Add lines to indicator

Maxim Korobov

Imagine indicator that loads news from data sources list, which was passed by params. List length is variable.

Now how to add dynamically lines to indicator?

backtrader

You cannot dynamically add lines. Not because it cannot be done, because the code filling the lines would also have to be changed dynamically and the number of lines would change along the datetime axis.

Maxim Korobov

Ok then is it possible to hide and rename lines? Thus one can add 10 lines with pretty generic names, then hide not necessary one and rename visible?

backtrader

It seems you are concerned about plotting those lines because you specifically say hide and thus the assumption that the same applies to rename.

You can change the plotlines information for any line.

See Docs - Plotting and specifically the section Line specific plotting options. The plotlines information can be changed in a live instance.

backtrader

From Stochastic for example

    lines = ('percK', 'percD',)
    ...

    plotlines = dict(percD=dict(_name='%D', ls='--'),
                     percK=dict(_name='%K'))

Used because % is not an allowed character in identifiers. And from the DrawDown observer

    lines = ('drawdown', 'maxdrawdown',)

    ...
 
    plotlines = dict(maxdrawdown=dict(_plotskip=True,))

Maxim Korobov

Here and in other indicators plotlines are static defined. Is _plotinit right place to redefine it?

backtrader

Definition only happens when the class is created. At any other time including _plotinit, things happen in instances and there you can only change the values of the already defined things.

Maxim Korobov

Got it too.
Last question: how to reach exact plotline in loop, not calling it by name? I.e. not like this

self.plotlines.<line_name>._name = 'new'

Studying AutoInfoClass...

backtrader

You cannot. What you can do is name the lines sequentially and then access them sequentially.

Maxim Korobov

Could you provide code snippet please?

backtrader

Standard python ...

lines = ('line1', ..., 'lineX')
plotlines = dict(
    line1=dict(...),
    ...
    lineX = dict(...),
)

for i in range(1, X + 1):
    plotline = getattr(self.plotlines, 'line' + i)
    # do something with the plotline

Maxim Korobov

Thank you for all replies and code snippets.

Here is the beauty:

alt text

I suggest you to add options to hide some text to make data cleaner - see red rectangles on image:

Indicator parameters list;
Line last value.

backtrader

Whatever is shown after the name of the indicator can be controlled. See the following excerpt from RSI:

    def _plotlabel(self):
        plabels = [self.p.period]
        plabels += [self.p.movav] * self.p.notdefault('movav')
        return plabels

The method returns a list of strings, with one special provision to use the name of some classes as the text label to automatically support using other indicators.

Maxim Korobov

Thanks!

It's much cleaner now:

cleaner

Dreaming of disable showing last value feature.

backtrader

Both _plotlabel and _plotinit have now been documented under Docs - Plotting

Maxim Korobov

Am I missing something or there is not option to reach (cancel) this code in Plot_OldSync.plotind()?

        if not math.isnan(lplot[-1]):
            label += ' %.2f' % lplot[-1]

backtrader

backtrader is not the 1st platform to plot ... and a common pattern is to plot the last value of the indicator. So far no option to disable it.

backtrader

This was anyhow not a breaking change, so the development branch contains now

Control over the last value after the name of the object
- Global: linevalues (as option to plot or in a PlotScheme instance)
- Per-object: plotinfo.plotlinevalues
- Per-line (for indicators/observers): _plotvalue
Control over the right hand side tag on lines:
- Global: valuetags (as option to plot or in a PlotScheme instance)
- Per-object: plotinfo.plotvaluetags
- Per-line (for indicators/observers): _plotvaluetag

backtrader

Part of the small release 1.9.38.116 and added to the plotting documentation

Maxim Korobov

In my particular case

line._plotvalue = False

helps.

Thanks for flexible solution!