https://github.com/chrisdone/labels/blob/master/labels-explore/src/Labels/Explore.hs#L26

@chrisdone This is only tangentially to do with labels. I was trying to emulate http://tomaugspurger.github.io/modern-1.html in Haskell. Using his data set which is 113M, I thought I would start off by seeing what the headers are. I ran this in ghci:

main2 = do
  foo <- fmap decodeByName (LB.readFile "examples/787338586_T_ONTIME.csv")
  case foo of
    Right (header, rows :: Vector (Map String String)) -> do
      putStrLn $ show header
    Left err -> putStrLn err

This took a long time to run and used over 6G of memory eventually returning

*Main> main2
["FL_DATE","UNIQUE_CARRIER","AIRLINE_ID","TAIL_NUM","FL_NUM","ORIGIN_AIRPORT_ID","ORIGIN_AIRPORT_SEQ_ID","ORIGIN_CITY_MARKET_ID","ORIGIN","ORIGIN_CITY_NAME","ORIGIN_STATE_NM","DEST_AIRPORT_ID","DEST_AIRPORT_SEQ_ID","DEST_CITY_MARKET_ID","DEST","DEST_CITY_NAME","DEST_STATE_NM","CRS_DEP_TIME","DEP_TIME","DEP_DELAY","TAXI_OUT","WHEELS_OFF","WHEELS_ON","TAXI_IN","CRS_ARR_TIME","ARR_TIME","ARR_DELAY","CANCELLED","CANCELLATION_CODE","DIVERTED","DISTANCE","CARRIER_DELAY","WEATHER_DELAY","NAS_DELAY","SECURITY_DELAY","LATE_AIRCRAFT_DELAY",""]

Is there a better way of doing this? I'd quite like to write a Haskell equivalent of a pandas tutorial. Maybe we should create a separate repo for doing this and tracking how it is going.

Note: You need GHC 8.0.1 for the #foo syntax, otherwise you have to use $("foo") which works on GHC 7.10

The lower base bound for 0.2 is base >= 4.9, so GHC 7.10 is rulled out by that

it can be implemented by add the new field name, and drop the old one.
but i want to know whether exist a more sensible way to do it

it seems the drop also can't be done directly, it need project's help

Currently, (:=) has a precedence of 6, which is too high if I want to do something like (#foo := x <$> y) because (<$>) has a precedence of only 4.

Is there any special reason the precedence of (:=) is not lower? Perhaps it should be 1?

The examples work and the code looks great, but the performance seems very poor.
Running
time stack exec labels-explore-example

gives on my laptop
flights distance
471949 3.7207249e8

real 0m54.148s
user 0m58.902s
sys 0m5.340s

The URLs in .gitmodule are of the form [email protected]:... instead of https://github.com/..., so git clone --recursive https://github.com/chrisdone/labels.git fails if the user has no GitHub account. To enable easy access, e.g. as part of an automated stack build using a robot user without any SSH key, the URLs in .gitmodule should be changed to the HTTPS form.

AFAIK, GitHub doesn't have any simple way to easily switch the view of the submodule URLs between HTTPS and SSH without changing the actual code of .gitmodule.

How can I easily get Generic instances for the Labels 'record' types? Generic is needed for deriving instances like Hashable and ToJSON.

If it can't be done easily, I suppose a workaround easily getting instances of things like Hashable is to use Labels.reflect and GHC.Generics.from to extract the field names and values Rep? Is there a better way?

Hi there was trying to follow your post 'Working with data in Haskell' https://www.fpcomplete.com/blog/2016/09/data-haskell

So decided to pull the labels-master package down from here and went into each directory and did the following

cabal install labels.cabal (ran successfully to completion)
cabal install labels-cvs.cabal (ran successfully to completion)
cabal install labels-json.cabal (ran successfully to completion)
cabal install zip-conduit.cabal (ran successfully to completion)
cabal install http-client.cabal (ran successfully to completion)

when I ran cabal install labels-explore.cabal (It fails, doesn't complete, the log says the following is wrong)

root@haskellbox~/.cabal/logs# cat labels-explore-0.1.0.0.log

cabal: Entering directory '.'
Configuring labels-explore-0.1.0.0...
Building labels-explore-0.1.0.0...
Preprocessing library labels-explore-0.1.0.0...
[1 of 1] Compiling Labels.Explore ( src/Labels/Explore.hs, dist/build/Labels/Explore.o )

src/Labels/Explore.hs:26:5: error:
Ambiguous occurrence ‘.|’
It could refer to either ‘Data.Conduit..|’,
imported from ‘Data.Conduit’ at src/Labels/Explore.hs:72:1-29
or ‘Labels.Explore..|’, defined at src/Labels/Explore.hs:112:1
cabal: Leaving directory '.'

Can you help please would like to get this built so i can run the examples thank you

I tried building labels with GHC 8.2 RC3 using this stack.yaml:

resolver: nightly-2017-07-20
compiler: ghc-8.2.0.20170704
compiler-check: match-exact
packages:
- labels

but got this error:

$ stack build labels
labels-0.3.2: build (lib)
Preprocessing library for labels-0.3.2..
Building library for labels-0.3.2..
[1 of 2] Compiling Labels.Internal  ( src/Labels/Internal.hs, .stack-work/dist/x86_64-linux/Cabal-2.0.0.0/build/Labels/Internal.o )

/home/jchia/gh/labels/labels/src/Labels/Internal.hs:66:5: error:
    • Couldn't match expected type ‘Proxy l'’
                  with actual type ‘p0 -> Proxy t0’
    • The equation(s) for ‘fromLabel’ have one argument,
      but its type ‘Proxy l'’ has none
      In the instance declaration for ‘IsLabel l (Proxy l')’
    • Relevant bindings include
        fromLabel :: Proxy l' (bound at src/Labels/Internal.hs:66:5)
   |
66 |     fromLabel _ = Proxy
   |     ^^^^^^^^^^^^^^^^^^^

--  While building package labels-0.3.2 using:
      /home/jchia/.stack/setup-exe-cache/x86_64-linux/Cabal-simple_mPHDZzAJ_2.0.0.0_ghc-8.2.0.20170704 --builddir=.stack-work/dist/x86_64-linux/Cabal-2.0.0.0 build lib:labels --ghc-options " -ddump-hi -ddump-to-file"
    Process exited with code: ExitFailure 1

I suspect it's because of GHC 8.2 changes to OverloadedLabels.

Unlike aeson's TH and Generics-derived instances, Maybe is not special cased by labels-json.

> data Foo = Foo { foo :: Maybe Int } deriving (Show, Generic, FromJSON)
> decode "{}" :: Maybe Foo
Foo { foo = Nothing }

> decode "{}" :: Maybe ("foo" := Maybe Int)
Nothing

the lables-csv only contains the FromJSON, not ToJSON.
so i wonder whether it is feasible to convert labels to json.

Chris, could you add ontime.csv.zip to this repo? It is no longer at http://chrisdone.com/ontime.csv.zip so examples cannot be run ... Thanks

currently i only can find cons function to expand record.
it's possible to merge two label record, and merge the same field?

*Main Labels> get #b (#a := 3, #b := "c", #a:="d")
"c"
*Main Labels> cons (#a := "d") (#a := 3)
(#a := "d",#a := 3)
*Main Labels> get #a (#a := 3, #b := "c", #a:="d")

<interactive>:65:1: error:
    • Overlapping instances for Has
                                  "a" value ("a" := value, "b" := value0, "a" := value)
      Matching givens (or their superclasses):
        Has "a" value ("a" := value, "b" := value1, "a" := value)
          bound by the inferred type for ‘it’:
                     forall value value1.
                     (Num value,
                      Has "a" value ("a" := value, "b" := value1, "a" := value),
                      Data.String.IsString value1, Data.String.IsString value) =>
                     value
          at <interactive>:65:1-36
      Matching instances:
        instance Has l a (l := a, u2, u3) -- Defined in ‘Labels.Internal’
        instance Has l a (u1, u2, l := a) -- Defined in ‘Labels.Internal’
    • In the ambiguity check for the inferred type for ‘it’
      To defer the ambiguity check to use sites, enable AllowAmbiguousTypes
      When checking the inferred type
        it :: forall value1 value2.
              (Num value1,
               Has "a" value1 ("a" := value1, "b" := value2, "a" := value1),
               Data.String.IsString value2, Data.String.IsString value1) =>
              value1