Skip to content

Commit

Permalink
Update Range to be a collection on contiguous regions; not finished.
Browse files Browse the repository at this point in the history
  • Loading branch information
Daniel Barowy committed Dec 31, 2015
1 parent 6cbdbc2 commit e6c7ed5
Show file tree
Hide file tree
Showing 3 changed files with 130 additions and 54 deletions.
170 changes: 118 additions & 52 deletions Parcel/AST.fs
View file
Expand Up @@ -2,6 +2,7 @@
open System
open System.Diagnostics
open System.Collections.Generic
open System.Text

type IndirectAddressingNotSupportedException(expression: string) =
inherit Exception(expression)
Expand All @@ -27,6 +28,19 @@

[<Serializable>]
type Address(R: int, C: int, env: Env) =
interface IComparable with
member self.CompareTo(obj) =
let addr = obj :?> Address
match env.WorksheetName.CompareTo(addr.WorksheetName) with
| -1 -> -1
| 1 -> 1
| _ ->
// this guarantees a consistent total order, but if used in user-facing
// contexts, it might not be very intuitive to end-users.
let c = Hash.cantorPair (System.Convert.ToUInt32 R) (System.Convert.ToUInt32 R) System.UInt32.MaxValue
let c' = Hash.cantorPair (System.Convert.ToUInt32 addr.Row) (System.Convert.ToUInt32 addr.Col) System.UInt32.MaxValue
c.CompareTo c'

static member fromR1C1(R: int, C: int, wsname: string, wbname: string, path: string) : Address =
Address(R, C, Env(path, wbname, wsname))
static member fromA1(row: int, col: string, wsname: string, wbname: string, path: string) : Address =
Expand Down Expand Up @@ -64,12 +78,12 @@
let k_1 = System.Convert.ToUInt32(R)
let k_2 = System.Convert.ToUInt32(C)
// get uint from pairing function
let pi = (((k_1 + k_2) * (k_1 + k_2 + 1u))/2u + k_2) % r
// shift pi depending on the worksheet 'index'
let pi = Hash.cantorPair k_1 k_2 r
// shift pi depending on the worksheet hashcode
// (we don't have access to the real index here; getting the
// sheet name's last digit is an approximation)
let lastLetter = (uint32 sheetname.[sheetname.Length - 1]) % w
let hashcode = pi + lastLetter * r
// sheet's hashcode mod w is an approximation)
let sheet_hc = (uint32 sheetname.[sheetname.Length - 1]) % w
let hashcode = pi + sheet_hc * r
int32 hashcode // this casts; it does not convert
// necessary because Address is used as a Dictionary key
override self.GetHashCode() : int =
Expand All @@ -85,13 +99,6 @@
let d = self.WorkbookName = addr.WorkbookName
let e = self.Path = self.Path
a && b && c && d && e
member self.InsideRange(rng: Range) : bool =
not (self.X < rng.getXLeft() ||
self.Y < rng.getYTop() ||
self.X > rng.getXRight() ||
self.Y > rng.getYBottom())
member self.InsideAddr(addr: Address) : bool =
self.X = addr.X && self.Y = addr.Y
override self.ToString() =
"(" + self.Y.ToString() + "," + self.X.ToString() + ")"
static member CharColToInt(col: string) : int =
Expand Down Expand Up @@ -129,56 +136,115 @@
// indirect references are essentially lambdas for
// constructing addresses
raise(IndirectAddressingNotSupportedException(expr))

and Range(topleft: Address, bottomright: Address) =
let _tl = topleft
let _br = bottomright

and Range(regions: (Address * Address) list) =
let _regions = regions
override self.ToString() =
let tlstr = topleft.ToString()
let brstr = bottomright.ToString()
tlstr + "," + brstr
let sregs = List.map (fun (tl, br) -> "(" + tl.ToString() + "," + br.ToString() + ")") _regions
String.Join(",", sregs)
member self.copyWithNewEnv(envnew: Env) =
Range(_tl.copyWithNewEnv(envnew), _br.copyWithNewEnv(envnew))
member self.TopLeft = _tl
member self.BottomRight = _br
Range(List.map (fun (tl: Address, br: Address) ->
tl.copyWithNewEnv(envnew), br.copyWithNewEnv(envnew)) _regions
)
member self.A1Local() : string =
_tl.A1Local() + ":" + _br.A1Local()
member self.getXLeft() : int = _tl.X
member self.getXRight() : int = _br.X
member self.getYTop() : int = _tl.Y
member self.getYBottom() : int = _br.Y
member self.InsideRange(rng: Range) : bool =
not (self.getXLeft() < rng.getXLeft() ||
self.getYTop() < rng.getYTop() ||
self.getXRight() > rng.getXRight() ||
self.getYBottom() > rng.getYBottom())
// Yup, weird case. This is because we actually
// distinguish between addresses and ranges, unlike Excel.
member self.InsideAddr(addr: Address) : bool =
not (self.getXLeft() < addr.X ||
self.getYTop() < addr.Y ||
self.getXRight() > addr.X ||
self.getYBottom() > addr.Y)
let sregs = List.map (fun (tl: Address, br: Address) ->
"(" + tl.A1Local() + "," + br.A1Local() + ")"
) _regions
String.Join(",", sregs)
member self.GetWorksheetNames() : seq<string> =
[_tl.WorksheetName; _br.WorksheetName] |> List.toSeq |> Seq.distinct
List.fold (fun wss (tl: Address, br: Address) ->
tl.WorksheetName :: br.WorksheetName :: wss
) [] _regions |>
List.toSeq |>
Seq.distinct
member self.GetWorkbookNames() : seq<string> =
[_tl.WorkbookName; _br.WorkbookName] |> List.toSeq |> Seq.distinct
List.fold (fun wss (tl: Address, br: Address) ->
tl.WorkbookName :: br.WorkbookName :: wss
) [] _regions |>
List.toSeq |>
Seq.distinct
member self.GetPathNames() : seq<string> =
[_tl.Path; _br.Path] |> List.toSeq
List.fold (fun wss (tl: Address, br: Address) ->
tl.Path :: br.Path :: wss
) [] _regions |>
List.toSeq |>
Seq.distinct
member self.Addresses() : Address[] =
Array.map (fun c ->
Array.map (fun r ->
Address.fromR1C1(r, c, _tl.WorksheetName, _tl.WorkbookName, _tl.Path)
) [|self.getYTop()..self.getYBottom()|]
) [|self.getXLeft()..self.getXRight()|] |>
Array.concat
// for every contigious region
List.map (fun (tl: Address, br: Address) ->
// for every column in that region
Array.map (fun c ->
// and every row in that region
Array.map (fun r ->
// get the address of the cell contained
Address.fromR1C1(r, c, tl.WorksheetName, tl.WorkbookName, tl.Path)
) [|tl.Y..br.Y|]
) [|tl.X..br.X|] |>
Array.concat
) _regions |>
Array.ofList |>
Array.concat |>
// ensure that we only enumerate overlapping cells once
Seq.distinct |>
Array.ofSeq

override self.GetHashCode() : int =
_tl.GetHashCode() ^^^ _br.GetHashCode()
Hash.jenkinsOneAtATimeHash _regions
override self.Equals(obj: obj) : bool =
let r = obj :?> Range
let a = topleft = r.TopLeft
let b = bottomright = r.BottomRight
a && b
let r_set = Set.ofArray(r.Addresses())
let self_set = Set.ofArray(self.Addresses())
self_set = r_set

// and Range(topleft: Address, bottomright: Address) =
// let _tl = topleft
// let _br = bottomright
// override self.ToString() =
// let tlstr = topleft.ToString()
// let brstr = bottomright.ToString()
// tlstr + "," + brstr
// member self.copyWithNewEnv(envnew: Env) =
// Range(_tl.copyWithNewEnv(envnew), _br.copyWithNewEnv(envnew))
// member self.TopLeft = _tl
// member self.BottomRight = _br
// member self.A1Local() : string =
// _tl.A1Local() + ":" + _br.A1Local()
// member self.getXLeft() : int = _tl.X
// member self.getXRight() : int = _br.X
// member self.getYTop() : int = _tl.Y
// member self.getYBottom() : int = _br.Y
// member self.InsideRange(rng: Range) : bool =
// not (self.getXLeft() < rng.getXLeft() ||
// self.getYTop() < rng.getYTop() ||
// self.getXRight() > rng.getXRight() ||
// self.getYBottom() > rng.getYBottom())
// // Yup, weird case. This is because we actually
// // distinguish between addresses and ranges, unlike Excel.
// member self.InsideAddr(addr: Address) : bool =
// not (self.getXLeft() < addr.X ||
// self.getYTop() < addr.Y ||
// self.getXRight() > addr.X ||
// self.getYBottom() > addr.Y)
// member self.GetWorksheetNames() : seq<string> =
// [_tl.WorksheetName; _br.WorksheetName] |> List.toSeq |> Seq.distinct
// member self.GetWorkbookNames() : seq<string> =
// [_tl.WorkbookName; _br.WorkbookName] |> List.toSeq |> Seq.distinct
// member self.GetPathNames() : seq<string> =
// [_tl.Path; _br.Path] |> List.toSeq
// member self.Addresses() : Address[] =
// Array.map (fun c ->
// Array.map (fun r ->
// Address.fromR1C1(r, c, _tl.WorksheetName, _tl.WorkbookName, _tl.Path)
// ) [|self.getYTop()..self.getYBottom()|]
// ) [|self.getXLeft()..self.getXRight()|] |>
// Array.concat
// override self.GetHashCode() : int =
// _tl.GetHashCode() ^^^ _br.GetHashCode()
// override self.Equals(obj: obj) : bool =
// let r = obj :?> Range
// let a = topleft = r.TopLeft
// let b = bottomright = r.BottomRight
// a && b

type ReferenceType =
| ReferenceAddress = 0
Expand Down
6 changes: 6 additions & 0 deletions Parcel/Grammar.fs
View file
Expand Up @@ -89,10 +89,16 @@
<|> AddrA1)

// Ranges

// contigious ranges
let MoreAddrR1C1 = pstring ":" >>. AddrR1C1
let RangeR1C1 = pipe2 AddrR1C1 MoreAddrR1C1 (fun r1 r2 -> Range(r1, r2))
let MoreAddrA1 = pstring ":" >>. AddrA1
let RangeA1 = pipe2 AddrA1 MoreAddrA1 (fun r1 r2 -> Range(r1, r2))
// discontiguous ranges
let DiscontRangeA1 = pipe2 RangeA1 RangeA1 (fun r1 r2 -> DRange(r1,r2))

//// top-level range
let RangeAny = (attempt RangeR1C1) <|> RangeA1

// Worksheet Names
Expand Down
8 changes: 6 additions & 2 deletions Parcel/Hash.fs
View file
Expand Up @@ -6,7 +6,7 @@
/// </summary>
/// <param name="key">A sequence of values of generic type.</param>
/// <returns>Returns a 32-bit integer.</returns>
let jenkinsOneAtATimeHash(key: seq<'a>) : uint32 =
let jenkinsOneAtATimeHash(key: seq<'a>) : int32 =
// https://en.wikipedia.org/wiki/Jenkins_hash_function#one-at-a-time
let hash = Seq.fold (fun hash elem ->
let hash1 = hash + uint32(elem.GetHashCode())
Expand All @@ -15,4 +15,8 @@ let jenkinsOneAtATimeHash(key: seq<'a>) : uint32 =
) 0u key
let hash1 = hash + (hash <<< 3)
let hash2 = hash1 ^^^ (hash1 >>> 11)
hash2 + (hash2 <<< 15)
let hash3 = hash2 + (hash2 <<< 15)
int32 hash3

let cantorPair(k_1: uint32)(k_2: uint32)(r: uint32) : uint32 =
(((k_1 + k_2) * (k_1 + k_2 + 1u))/2u + k_2) % r

0 comments on commit e6c7ed5

Please sign in to comment.