This library contains some useful functions for using the map-reduce-folds package with Frames (containers of data rows) from the Frames package. Included, in Frames.MapReduce, are helpers for filtering Frames, splitting records into key and data columns and reattaching key columns after reducing.
There is also support for using the map-reduce-folds library with
Vinyl records of more complex type, e.g. ARec (Maybe :. ElField) rs
.
The specific case of Maybe
is handled in Frames.MapReduce.Maybe and
the fully polymorphic case (supporting Rec
, ARec
and SRec
and any interpretation functor composed with ElField which has a reasonable
interpretation as Maybe
, e.g., Either
) is handled in
Frames.MapReduce.General.
Also included, in the Frames.Folds (and Frames.Folds.Maybe, Frames.Folds.General) module, are some helpful functions for building folds of Frames from folds over each column, specified either individually or via a constraint on all the columns being folded over.
There is a set of modules for aggregations
(Frames.Aggregation, Frames.Aggregation.General, Frames.Aggregation.Maybe),
a common map-reduce pattern where a set of keys (ages, locations, e.g.,)
are merged to some smaller set of keys (Child/Adult, Urban/Rural, e.g,)
and the data columns (e.g., population, fraction vaccinated against flu) can be
aggregated to data for the merged group.
Given a function from old keys to new keys (as records) and a fold over
the data expressing the aggregation, this function will build the fold to do the
aggregation over the entire data-set.
There are helper functions to:
- Build the data fold from folds for each field.
- Lift functions on types (Age -> Child/Adult) to functions on record columns containing those types.
- Merge key functions (Age -> Child/Adult, Location -> Urban/Rural) -> ((Age,Location) -> (Child/Adult,Urban/Rural))
NB: The functions which operate on Record rs
, record (Maybe :. ElField) rs
and Applicative f => record (f :. ElField)
, have the same names but reside in
different modules. I assume most users will want only one version. But if, as in the
example provided, you want to use two versions in the same module you will need to import
at least one of them qualified.
For example, given a Frame with three columns, a text column Label
and two columns, X
and Y
, holding doubles, we
-
unpack, filtering using
unpackFilterOnField
(with a type-application to specify theLabel
column), -
assign, thus grouping by
Label
and feeding the rest of the columns to reduce usingsplitOnKeys
with a type-application to specify which columns are the key. -
reduce by folding over the two remaining columns using the
foldAllConstrained
function. The type-application here specifies a constraint satisfied by all the columns being folded, and then the cols to fold.
This last part is a little complex. See the Frames.Folds modules for more details.
We can use an aggregation:
- Aggregate the data using a sum on the
Y
field and a Y-weighted sum on theX
field. - Create an aggregate key, in this case one which represents the label being "A" or "B" vs. anything else
- Create a function to map records of Labels into records of our new key.
- Build the resulting fold, using
aggregateFold
. NB: If we had more key columns we could hold them constant and we would need specify them via type-application toaggregateFold
. In this case we have no such columns so that argument is'[]
We can also demonstrate the functions operating on Rec (Maybe :. ElField)
rows.
- unpack, doing nothing.
- assign as above, using the
Maybe
generalized version ofsplitOnKeys
. - reduce as above, using the
Maybe
generalized versions offoldAndAddKey
andfoldAllConstrained
.
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE OverloadedStrings #-}
{-# LANGUAGE TypeApplications #-}
{-# LANGUAGE TypeOperators #-}
{-# LANGUAGE InstanceSigs #-}
module Main where
import qualified Control.Foldl as FL
import qualified Data.List as L
import qualified Data.Text as T
import qualified Data.Vinyl as V
import Data.Vinyl.Functor ( Compose(..)
, (:.)
)
import qualified Frames as F
import qualified Frames.CSV as F
import qualified Frames.Folds as FF
import qualified Frames.Folds.Maybe as FFM
import qualified Frames.MapReduce as FMR
import qualified Frames.MapReduce.Maybe as FMRM
import System.Random ( newStdGen
, randomRs
)
-- Create types for the cols
type Label = "label" F.:-> T.Text
type Y = "y" F.:-> Double
type X = "x" F.:-> Double
type AllCols = '[Label, Y, X]
-- filter, leaving only rows with labels 'A', 'B' or 'C'
unpack = FMR.unpackFilterOnField @Label (`elem` ["A", "B", "C"])
-- group the rest of the cols by Label
assign = FMR.splitOnKeys @'[Label]
-- sum the data columns and then re-attach the key
reduce = FMR.foldAndAddKey $ (FF.foldAllConstrained @Num @'[Y, X]) FL.sum
-- put it all together: filter, group by label, sum the data cols and re-attach the key.
-- Then turn the resulting list of Frames (each with only one Record in this case)
-- into one Frame via (<>).
mrFold = FMR.concatFold $ FMR.mapReduceFold unpack assign reduce
aggDataFold :: FL.Fold (F.Record '[Y, X]) (F.Record '[Y, X])
aggDataFold =
let sumYF = FL.premap (F.rgetField @Y) FL.sum
sumProdXYF = FL.premap (\r -> F.rgetField @X r * F.rgetField @Y r) FL.sum
wgtdSumXF = (\sXY sY -> sXY / sY) <$> sumProdXYF <*> sumYF
in FF.sequenceRecFold
$ FF.toFoldRecord sumYF
V.:& FF.toFoldRecord wgtdSumXF
V.:& V.RNil
data AggKey = AorB | Other deriving (Eq, Ord, Show)
type instance FI.VectorFor AggKey = Vec.Vector
type AggKeyCol = "AggKey" F.:-> AggKey
groupLabels :: FA.RecordKeyMap '[Label] '[AggKeyCol]
groupLabels = FA.keyMap $ \l -> if (l `elem` ["A", "B"]) then AorB else Other
aggFold = FA.aggregateFold @'[] groupLabels aggDataFold
-- Bleh, this should go in Frames.
instance (Eq (F.ElField a)) => Eq (Compose Maybe F.ElField a) where
(==) (Compose fga) (Compose fga') = fga == fga'
instance (Ord (F.ElField a)) => Ord (Compose Maybe F.ElField a) where
compare (Compose fga) (Compose fga') = fga `compare` fga'
unpack'
:: FMR.Unpack (F.Rec (Maybe :. F.ElField) rs) (F.Rec (Maybe :. F.ElField) rs)
unpack' = FMRM.unpackNoOp
assign'
:: FMR.Assign
(F.Rec (Maybe :. F.ElField) '[Label])
(F.Rec (Maybe :. F.ElField) '[Label, X, Y])
(F.Rec (Maybe :. F.ElField) '[X, Y])
assign' = FMRM.splitOnKeys @'[Label]
reduce'
:: FMR.Reduce
(F.Rec (Maybe :. F.ElField) '[Label])
(F.Rec (Maybe :. F.ElField) '[X, Y])
(F.Rec (Maybe :. F.ElField) '[Label, X, Y])
reduce' = FMRM.foldAndAddKey $ (FFM.foldAllConstrained @Num @'[X, Y]) FL.sum
mrFold'
:: FMR.Fold
(F.Rec (Maybe :. F.ElField) '[Label, X, Y])
[F.Rec (Maybe :. F.ElField) '[Label, X, Y]]
mrFold' = FMR.mapReduceFold unpack' assign' reduce'
main :: IO ()
main = do
f <- createFrame 1000
let result = FMR.fold mrFold f
putStrLn $ (L.intercalate "\n" $ fmap show $ FL.fold FL.list result)
let result' = FMR.fold mrFold' createHolyRows
putStrLn . unlines . fmap show $ FL.fold FL.list result'
let aggregatedResult = FMR.fold aggFold f
putStrLn $ (L.intercalate "\n" $ fmap show $ FL.fold FL.list aggregatedResult)
{- Output
{label :-> "A", y :-> 1577.3965303339942, x :-> 1507.286289962377}
{label :-> "B", y :-> 1934.223021597267, x :-> 2135.9312483902577}
{label :-> "C", y :-> 1528.6898777108415, x :-> 1810.5096765228654}
{Just label :-> "A", Just x :-> 5.0, Just y :-> 2.0}
{Just label :-> "Z", Just x :-> 5.0, Just y :-> 9.0}
{AggKey :-> AorB, y :-> 3857.3338804158475, x :-> 48.675203593420946}
{AggKey :-> Other, y :-> 45655.25138686513, x :-> 47.24947893453326}
-}
--- create the Frame
createFrame :: Int -> IO (F.FrameRec AllCols)
createFrame n = do
g <- newStdGen
let randLabels = L.take n $ randomRs ('A', 'Z') g
randDbls = L.take (2 * n) $ randomRs (0.0, 100.0) g
oneRow m =
T.singleton (randLabels !! m)
F.&: (randDbls !! m)
F.&: (randDbls !! (n + m))
F.&: V.RNil
return $ F.toFrame $ fmap oneRow [0 .. (n - 1)]
createHolyRows :: [F.Rec (Maybe F.:. F.ElField) '[Label, X, Y]]
createHolyRows = fmap go [one, two, three, four]
where
go =
V.rmap (either (const (Compose Nothing)) (Compose . Just) . getCompose)
. F.readRec
one = ["A", "1", "2"]
two = ["Z", "NaN", "3"]
three = ["A", "4", "lol"]
four = ["Z", "5", "6"]
LICENSE (BSD-3-Clause)
Copyright (c) 2018, Adam Conner-Sax, All rights reserved.
Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above
copyright notice, this list of conditions and the following
disclaimer in the documentation and/or other materials provided
with the distribution.
* Neither the name of Adam Conner-Sax nor the names of other
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.