Mathias Brandewinder on .NET, F#, VSTO and Excel development, and quantitative analysis / machine learning.
6. February 2013 15:06

In spite of being color blind, I am a visual guy – I like to see things. Nothing beats a chart to identify problems in your data. I also spend lots of time manipulating data in FSI, the F# REPL, and while solutions like FSharpChart makes it possible to produce nice graphs fairly easily, I still find it introduces a bit of friction, and wondered how complicated it would be to use Excel as a charting engine.

Turns out, it’s not very complicated. The typical use case for generating charts in Excel is to first put data in a spreadsheet, and use the corresponding range as a source for a chart. However, it’s also perfectly possible to directly create a Chart object, and manipulate its SeriesCollection, adding and editing Series, which are arrays of XValues and Values.

As a starting point, I decided to focus on 2 problems:

• plotting functions, in 2 and 3 dimensions,
• producing scatterplots.

Both are rather painful to do in Excel itself – and scatterplots are the one chart I really care about when analyzing data, because it helps figuring out whether or not some variables are related.

What I wanted was a smooth experience from FSI – start typing code, and ship data to Excel, without having to worry about the joys of the Excel interop and its syntax. The video below shows what I ended up with, in action.

Note: watching me type is about as exciting as watching paint dry, so I sped up the video from its original 5 minutes down to 2 - otherwise there is no trick or editing.

This year’s blockbuster: plotting functions from F# to Excel

I’ll try to do another one on scatterplots later. In the meanwhile, here are some comments on the script, which you can find here on GitHub.

I really wanted to shield the user from dealing with Excel interop (if you have had the pleasure to deal with it, you know why) – the two functions below help achieving that:

// Attach to the running instance of Excel, if any
let Attach () =
try
Marshal.GetActiveObject("Excel.Application")
:?> Microsoft.Office.Interop.Excel.Application
|> Some
with
| _ ->
printfn "Could not find running instance of Excel"
None

// Find the Active workbook, if any
let Active () =
let xl = Attach ()
match xl with
| None -> None
| Some(xl) ->
try
xl.ActiveWorkbook |> Some
with
| _ ->
printfn "Could not find active workbook"
None


The first looks for a running instance of Excel, and “attaches” to it, and the second finds the currently active workbook, where new Charts will be produced. As a result, as long as Excel is open, the script will know where to do its work.

Caveat: if multiple instances of Excel are open (which is typically not the case), results might be a bit unpredictable.

I went back and forth, but ended up implementing the function plot as a Class, because maintaining some state simplified quite a bit things like adding functions to an existing plot, and resizing / zooming. Here is the full code for Plot, with some comments afterwards:

type Plot (f: float -> float, over: float * float) =
let mutable functions = [ f ]
let mutable over = over
let mutable grain = 50
let chart = NewChart ()
let values () =
let min, max = over
let step = (max - min) / (float)grain
[| min .. step .. max |]
let draw f =
match chart with
| None -> ignore ()
| Some(chart) ->
let seriesCollection = chart.SeriesCollection() :?> SeriesCollection
let series = seriesCollection.NewSeries()
let xValues = values ()
series.XValues <- xValues
series.Values <- xValues |> Array.map f
let redraw () =
match chart with
| None -> ignore ()
| Some(chart) ->
let seriesCollection = chart.SeriesCollection() :?> SeriesCollection
for s in seriesCollection do s.Delete() |> ignore
functions |> List.iter (fun f -> draw f)

do
match chart with
| None -> ignore ()
| Some(chart) ->
chart.ChartType <- XlChartType.xlXYScatter
let seriesCollection = chart.SeriesCollection() :?> SeriesCollection
draw f

member this.Add(f: float -> float) =
match chart with
| None -> ignore ()
| Some(chart) ->
functions <- f :: functions
draw f

member this.Rescale(min, max) =
over <- (min, max)
redraw()

member this.Zoom(zoom: int) =
grain <- zoom
redraw()   

Plot maintains a list of functions with signature float –> float, an interval (a tuple of floats) over which to plot them, and a “grain”, which represents how many points will be plotted over that interval. The values() function generates the X-values of the chart, by dividing equally the interval proportionally to the grain, and draw f adds a new Series to the chart, filling in the XValues with values(), and mapping each of them by the function f. Three methods are publicly exposed: Add (to add a new function to the Plot), Rescale (to change the bounds of the display interval), and Zoom (to set the granularity of the display).

The usage of Plot is as simple as the following: load the script into FSI, launch Excel, and go:

> let f x = cos x;;

val f : x:float -> float

> let plot = Plot(f, (0., 1.));;

val plot : Plot

> let pi = System.Math.PI;;

val pi : float = 3.141592654

> plot.Rescale(-pi, pi);;

val it : unit = ()

> plot.Zoom(200);;

val it : unit = ()

The Surface plot is very similar, except that it expects a function of 2 arguments, like this:

> let g x y = cos x * sin y;;

val g : x:float -> y:float -> float

> let s = Surface(g, (0., pi), (0., pi));;

val s : Surface

That’s it for today! Next time, I’ll talk about the scatterplots. In the meanwhile, if you have feedback, I’d love to hear it. This is still work in progress (obviously, I need to add classic charts like histograms, bars, and lines, this is coming soon), and I am designing it for my own needs – if you see something which would make it better for you, let me know  - or place a pull request!

## Resources

Excel-Charts: full script on GitHub

FSharpChart: a F# library to generate charts from FSI.

Unconstrained optimization test functions: Dr. Abdel-Rahman Hedar awesome collection of funky functions.

4/8/2013 3:02:45 AM # Cool!

1. You're defining a "flatten" and a "Selection" function there that you don't use in the rest of your class and that could be left out (I imagine you might use them in FSI sessions, but when refactoring to an API you don't need them).

2. You could refactor the code a little by defining (within the Plot class):

let applyToChart = fun cont ->
match chart with
| None ->  ignore()
| Some(chart) -> cont chart

3. Setting some of the excel settings and attributes (screenupdating, chart titles etc) could be factored out as functions

4/8/2013 4:45:51 AM # Hi Bram,
Thanks for the feedback! The flatten and Selection are indeed there for FSI sessions, so that you can grab the active selection and, for instance, transform it into a Scatterplot. I haven't had time to put together an example usage yet, but I'll do that in some hopefully not-to-far future.
There are definitely some rough edges in the script at that point, it's probably visible that I oscillated between objects and functions. I like your suggestions, and will look into it. I focused on the charts that are the hardest to produce from "standard" Excel (scatterplots and surfaces), but I want to add also classics, so some refactoring will have to take place anyways!
Have a great day,
Mathias

8/23/2013 11:30:10 PM # ça ne fonctionne pas avec Excel 2012 8/24/2013 4:43:32 AM # Damn - that is a bit strange, I will look into it. Thank you for letting me know!
Mathias   