How do I make a bar chart with Kandy for this Data...

# datascience

[dave08]

How do I make a bar chart with Kandy for this DataFrame:

VERSION: String
NAMESPACE:
    dev: Int
    istio-system: Int
    loki: Int
    monitor: Int
    prod: Int
    redis: Int

More in the thread

[dave08]

The original table was:

NAMESPACE: String
VERSION: String

and I did this on it:

psVersions.pivot { NAMESPACE }.groupBy { VERSION }.count()

[dave08]

I tried:|

versionStats.plot { 
    bars { 
        x(VERSION)
        y(NAMESPACE.values())
    }
}

but I got a very unclear error:

The problem is found in one of the loaded libraries: check library renderers
java.lang.IllegalArgumentException: Can't serialize object { dev:32, istio-system:0, lokistats:0, monitor:0, prod:0, redis:0 }
org.jetbrains.kotlinx.jupyter.exceptions.ReplLibraryException: The problem is found in one of the loaded libraries: check library renderers

[Andrei Kislitsyn]

Hi! What do you expect the y values of bars to be?

[dave08]

the count

[Andrei Kislitsyn]

The count of what value?

[dave08]

psVersions.pivot { NAMESPACE }.groupBy { VERSION }.count()

[dave08]

count()

[dave08]

I basically have a list of pods in all my namespaces, each pod has an istio version, I'd like to count how many pods of each version there is in each namespace

[dave08]

There's only two or three possible versions at a time

[dave08]

So X is the namespace, with a bar in each namespace for each version, and the count of pods in each of them is Y

[Andrei Kislitsyn]

Please, try (with original dataframe)

psVersions.groupBy {NAMESPACE} plot {
  countPlot(VERSION)
}

Is that what you expect?

[dave08]

btw plot is red with 'infix' modifier is required on 'plot' in 'org. jetbrains. kotlinx. kandy. dsl'

[dave08]

Nope, that gives me (not using infix) the namespaces in each version, I want the versions in each namespace

[dave08]

]

[Andrei Kislitsyn]

Oh, got it, then

psVersions.groupBy {VERSION} plot {
  countPlot(NAMESPACE)
}

should work for you.

[Andrei Kislitsyn]

Is it?😄

[dave08]

Wow so simple... I didn't really expect that Kandy could plot with all that extra data still there, I really thought I had to eliminate all the data except for Kandy's input.

[dave08]

I looked all over the docs, and couldn't find a hint on this one.

[dave08]

Now that I know it seems logical, since you actually need to specify the appropriate fields.

[Andrei Kislitsyn]

Count plot is described here:
https://kotlin.github.io/kandy/count-plot-guide.html

[dave08]

Yeah, that's for two dimensions... when I started with pivots I was already getting confused how to do it.

[Andrei Kislitsyn]

You can do the same with your pivoted dataframe, but more complicated - there is one problem described here (https://kotlin.github.io/kandy/series-hack-guide.html) why Kandy can’t work with columns in “pyplot style”.

[Andrei Kislitsyn]

So you need to gather (https://kotlin.github.io/dataframe/gather.html) NAMESPACE subcolumns into one column (“count”) with new key column (“namespace”) and write something like that:

df.plot {
  bars {
     x(namespace)
     y(count)
     fillColor(version)
  }
}

[Andrei Kislitsyn]

But countPlot solution is a way better, of course.

[dave08]

Yeah, I guess gather is just trying to "depivot" it... It's still a bit obscure how countPlot uses NAMESPACE to count the pods in it...?

[dave08]

Oh... it's just the count of rows for each namespace 😅!

[Andrei Kislitsyn]

Yep 😁

[dave08]

Thanks for all the help!