Skip to content

deemp/clerk

Repository files navigation

clerk

clerk provides a Haskell library for declaratively describing the spreadsheets.. clerk is built on top of the xlsx package and extends upon the work of Nickolay Kudasov.

Features

clerk can be used to produce a styled spreadsheet with some data and formulas on it. These formulas are evaluated when the document is loaded into a target spreadsheet system.

The library supports:

  • Typed cell references. Example: Ref Double.
  • Type-safe arithmetic operations with them. Example: (a :: Ref Double) + (b :: Ref Double) produces a Ref Double.
  • Constructing expressions with given types. Example: (e :: Expr Double) = "SUM" [a .: b], e translates to SUM(A1:B1) (actual value depends on the values of a and b).
  • Conditional styles, formatting, column widths.

The examples below demonstrate these features.

Example 1

The goal: demonstrate formula syntax.

The source code for this example is available here.

Extensions

{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE TypeApplications #-}
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE ScopedTypeVariables #-}

Imports

I import the necessary stuff.

import Clerk ( (.*), (.**), (.+), (./), as, fun, formulaRef,  val, Formula, Ref )
import Data.Text (Text)
import Examples.Helpers ( showFormula, mkRef )

Formulas

Formulas consist of references, functions, and values.

I make references to Double values

r1 :: Ref Double
r1 = mkRef @"B4"
r2 :: Ref Double
r2 = mkRef @"E6"
r3 :: Ref Double
r3 = mkRef @"G8"

Next, I convert one of these references to a formula via formulaRef and inspect the formula representation.

t1 :: Text
t1 = showFormula $ formulaRef r2

-- >>>t1
-- "E6"

Finally, I construct a longer expression and look at its representation. I convert a literal value to a formula via val.

t2 :: Text
t2 = showFormula $ r1 .* r2 .* val 3 .+ r1 .** r2 ./ r3

-- >>>t2
-- "B4*E6*3.0+B4^E6/G8"

Of course, I can mix differently typed references in expressions when necessary. For this case, I have an unsafe as function.

r4 :: Ref Int
r4 = mkRef @"T6"

t3 :: Text
t3 = showFormula $ as @Double (r4 .* r4) .+ r1 .** r2 ./ r3

-- >>>t3
-- "T6*T6+B4^E6/G8"

This as function should not be abused, though. If I need an Int instead of a Double, I can explicitly use an Excel function.

round_ :: forall a. Formula a -> Formula Int
round_ x = fun "ROUND" [x]

t4 :: Formula Int
t4 = round_ (r1 .** r2 ./ r3)

-- >>>:t t4
-- t4 :: Formula Int

t5 :: Text
t5 = showFormula t4

-- >>> t5
-- "ROUND(B4^E6/G8)"

Example 2

The goal: describe and generate a spreadsheet with a simple multiplication table.

The source code for this example is available here.

The program produces an xlsx file that looks as follows:

With formulas enabled:

The below sections describe how such a spreadsheet can be constructed.

Extensions

I'll need several language extensions.

Imports

I import the necessary stuff.

import Clerk
import Control.Monad (forM, forM_, void)
import qualified Data.Text as T
import Lens.Micro ((&), (+~), (^.))

Tables

The tables that I'd like to construct are:

  • A vertical header
  • A horizontal header
  • A table with results of multiplication of the numbers from these headers

A vertical header

clerk provides the RowI monad. This monad takes some input, internally converts it into Excel types, and outputs something, e.g., a cell reference. In background, it writes a template of a horizontal block of cells - a row. This row is used for placing the input values onto a sheet.

A vertical block of cells can be represented as several horizontal blocks of cells placed under each other. So, as a template, I use a RowI with one integer as an input.

As I don't need any formatting, I use blank cells for templates.

I place the rows for each input value and collect the references. Each row is shifted relative to the input coordinates.

mkVertical :: Coords -> [Int] -> Sheet [Ref Int]
mkVertical coords numbers =
  forM (zip [0 ..] numbers) $ \(idx, number) ->
    placeIn
      (coords & row +~ idx + 2)
      number
      ((columnF blank (const number)) :: RowI Int (Ref Int))

A horizontal header

For a horizontal header, I make a row of numbers and collect the references to all its cells. As I don't care about the type of inputs, I use the Row type.

In the Sheet monad, I place this row starting at a specified coordinate.

mkHorizontal :: Coords -> [Int] -> Sheet [Ref Int]
mkHorizontal coords numbers =
  place
    (coords & col +~ 2)
    ((forM numbers $ \n -> columnF blank (const n)) :: Row [Ref Int])

Table builder

For inner cells, I use single-cell rows for each input. As I don't need any info about these cells, I use the Row () type.

mkTable :: [(Ref Int, Ref Int)] -> Sheet ()
mkTable cs =
  forM_ cs $ \(r, c) -> do
    coords <- mkCoords' (c ^. col) (r ^. row)
    place coords ((columnF_ blank (const (r .* c))) :: Row ())

Sheet

Now, I combine all functions.

sheet :: Sheet ()
sheet = do
  start <- mkCoords @"B2"
  let numbers = [1 .. 9]
  cs <- mkHorizontal start numbers
  rs <- mkVertical start numbers
  mkTable [(r, c) | r <- rs, c <- cs]

Result

Finally, I can write the result and get a spreadsheet like the one at the beginning of Example 2.

main :: IO ()
main = writeXlsx "example2.xlsx" [(T.pack "List 1", void sheet)]

To get ./example2.xlsx, run:

nix run .#example2
-- or
cabal run example2

With formulas enabled, the sheet looks like this:

Example 3

The goal: describe and generate a spreadsheet that calculates the pressure data given some volume data and constants.

The source code for this example is available here.

The program produces an xlsx file that looks as follows:

With formulas enabled:l

The below sections describe how such a spreadsheet can be constructed.

Extensions

I'll need several language extensions.

-- to access the fields of records like a.b
{-# LANGUAGE OverloadedRecordDot #-}
{-# LANGUAGE ImportQualifiedPost #-}
{-# LANGUAGE RankNTypes #-}
{-# LANGUAGE RecordWildCards #-}
{-# LANGUAGE DuplicateRecordFields #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE NoOverloadedStrings #-}
{-# LANGUAGE TypeApplications #-}
{-# LANGUAGE DataKinds #-}

Imports

And import the necessary stuff.

import Clerk
import Codec.Xlsx qualified as X
import Codec.Xlsx.Formatted qualified as X
import Data.Text qualified as T
import Lens.Micro ((%~), (&), (+~), (?~))

Tables

The tables that I'd like to construct are:

  • A table per a constant's value (three of them)
  • A volume and pressure table
  • A constants' header
  • A volume and pressure header

constants values

In our case, each constant has the same type of the numeric value - Double. However, it might be the case that in another set of constants, they'll have different types. That's why, I'll construct a table with a single row per a constant and later place the constants' tables under each other. I'll store constant data in a record.

data ConstantData a = ConstantData
  { constantName :: String
  , constantSymbol :: String
  , constantValue :: a
  , constantUnits :: String
  }

Next, I group the constants.

data Constants f = Constants
  { gasConstant :: f Double
  , numberOfMoles :: f Double
  , temperature :: f Double
  }

type ConstantsInput = Constants ConstantData

Following that, I record the constants data.

constants :: ConstantsInput
constants =
  Constants
    { gasConstant = ConstantData "GAS CONSTANT" "R" 0.08206 "L.atm/mol.K"
    , numberOfMoles = ConstantData "NUMBER OF MOLES" "n" 1 "moles"
    , temperature = ConstantData "TEMPERATURE(K)" "T" 273.2 "K"
    }

Now, I can make a RowI for a constant input. I use a RowI because this row cares about the input type. I'll later use this row for each constant separately.

I get a pair of outputs:

  • Top left cell of a constant's table. That is, the cell with that constant's name.
  • The value of the constant.

Later, I'll use these outputs to relate the positions of tables on a sheet.

Notice that I use styles like lightBlue here. These styles are defined in the Styles section.

constant :: (ToCellData a) => RowI (ConstantData a) (Ref (), Ref a)
constant = do
  refTopLeft <- columnF lightBlue constantName
  columnF_ lightBlue constantSymbol
  refValue <- columnF (lightBlue .& with2decimalDigits) constantValue
  columnF_ lightBlue constantUnits
  return (refTopLeft, refValue)

volume and pressure values

To fill this table, I'll take some data and combine it with the constants.

newtype Volume = Volume {volume :: Double}

volumeData :: [Volume]
volumeData = Volume <$> [1 .. 10]

To pass the constants references in a structured way, I make a helper type.

type ConstantsRefs = Constants Ref

Next, I define a function to produce a row for volume and pressure.

values :: ConstantsRefs -> RowI Volume ()
values Constants{..} = do
  refVolume <- columnF alternatingColors volume
  let pressure' = gasConstant .* numberOfMoles .* temperature ./ refVolume
  columnF_ (alternatingColors .& with2decimalDigits) (const pressure')

Constants' header

I won't use records here. Instead, I'll put the names of the columns straight into the Row.

The outputs will be the coordinates of the top left cell and the top right cell of this table.

constantsHeader :: Row (Ref (), Ref ())
constantsHeader = do
  let style :: FormatCell
      style = blue .& alignedCenter
  refTopLeft <- columnWF 20 style (const "constant")
  columnWF_ 8 style (const "symbol")
  columnF_ style (const "value")
  refTopRight <- columnWF 13 style (const "units")
  return (refTopLeft, refTopRight)

Volume & Pressure header

For this header, I'll also put the names of columns straight into a row.

valuesHeader :: Row (Ref ())
valuesHeader = do
  refTopLeft <- columnWF 12 green (const "VOLUME (L)")
  columnWF_ 16 green (const "PRESSURE (atm)")
  return refTopLeft

Sheet builder

At last, I combine all rows.

sheet :: Sheet ()
sheet = do
  start <- mkCoords @"B2"
  (constantsHeaderTL, constantsHeaderTopRight) <- place start constantsHeader
  (gasTopLeft, gas) <- placeIn (constantsHeaderTL & row +~ 2) constants.gasConstant constant
  (nMolesTopLeft, nMoles) <- placeIn (gasTopLeft & row +~ 1) constants.numberOfMoles constant
  temperature <- snd <$> placeIn (nMolesTopLeft & row +~ 1) constants.temperature constant
  valuesHeaderTopLeft <- place (constantsHeaderTopRight & col +~ 2) valuesHeader
  placeIns (valuesHeaderTopLeft & row +~ 2) volumeData (values $ Constants gas nMoles temperature)

Styles

I used several styles to format the tables. This is how these styles are defined.

blue, lightBlue, green, lightGreen :: FormatCell
blue = mkColor (hex @"#FF99CCFF")
lightBlue = mkColor (hex @"#90CCFFFF")
green = mkColor (hex @"#FF00FF00")
lightGreen = mkColor (hex @"#90CCFFCC")

alternatingColors :: FormatCell
alternatingColors index = (if even index then lightGreen else lightBlue) index

Additionally, I compose an FCTransform for the number format. Such a transform is used to accumulate cell formatting.

with2decimalDigits :: FCTransform
with2decimalDigits fcTransform =
  fcTransform & X.formattedFormat %~ X.formatNumberFormat ?~ X.StdNumberFormat X.Nf2Decimal

And I make a transform for centering the cell content.

alignedCenter :: FCTransform
alignedCenter = horizontalAlignment X.CellHorizontalAlignmentCenter

Result

Finally, I write the result and get the spreadsheet like the one at the beginning of Example 3.

main :: IO ()
main = writeXlsx "example3.xlsx" [(T.pack "List 1", sheet)]

To get ./example3.xlsx, run:

nix run .#example3
-- or
cabal run example3

With formulas enabled, the sheet looks like this:

{-# LANGUAGE DataKinds #-}
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE TypeApplications #-}

import Clerk
import Control.Monad (void)
import Control.Monad.RWS (gets)

main :: IO ()
main = writeXlsx "example4.xlsx" [("List 1", sheet 9 15)]

colFormula :: ToCellData output => output -> RowI input (Ref a)
colFormula = columnF blank . const

colIndex :: InputIndex -> RowI input ()
colIndex = void . colFormula

index :: Row InputIndex
index = gets ((+ 1) . _inputIndex)

row0 :: Int -> Int -> Row (Ref Int, Ref Int)
row0 a b = do
  colIndex =<< index
  r1 <- colFormula a
  r2 <- colFormula b
  pure (r1, r2)

row1 :: (Ref Int, Ref Int) -> Row (Ref Int, Ref Int)
row1 (a, b) =
  do
    colIndex =<< index
    r1 <- colFormula (fun "MAX" [a, b] :: Formula Int)
    r2 <- colFormula (fun "MIN" [a, b] :: Formula Int)
    pure (r1, r2)

row3 :: (Ref Int, Ref Int) -> Row (Ref Int, Ref Int)
row3 (a, b) = do
  colIndex =<< index
  r1 <- colFormula (formulaRef a)
  r2 <- colFormula (formulaRef b)
  r3 <- colFormula (fun "MOD" [r1, r2] :: Formula Int)
  pure (r2, r3)

sheet :: Int -> Int -> Sheet ()
sheet a b = do
  start <- mkRef @"A1"
  s1 <- place start (row0 a b)
  placeIxsFs_ start [1 :: Int .. 6] (cycle [row1, row3]) s1
  pure ()

Contribute

This project provides a dev environment via a Nix flake.

  1. With flakes enabled, run:

    nix develop
    cabal build
  2. This README.md is generated from several files. If you edit them, re-generate it.

    cabal test docs
  3. (Optionally) Start VSCodium with Haskell extensions.

    1. Write settings and run VSCodium.

      nix run .#writeSettings
      nix run .#codium .
    2. Open a Haskell file. Haskell Language Server should soon start giving you hints.

  4. Study these links if you'd like to learn more about the tools used in this flake:

  5. If Haskell Language Server doesn't want to run code in -- >>> comments:

    1. Check the Output of HLS.
    2. Find there a directory name containing hie-bios.
    3. Remove the hie-bios/dist-clerk-* directory.

About

Declaratively generate spreadsheets

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •