[Repo Assist] Add PreferFloats static parameter to CsvProvider (#1655)

* Add PreferFloats static parameter to CsvProvider

Adds a new boolean static parameter `PreferFloats` (default: `false`) to
`CsvProvider` that causes the type inference to use `float` instead of
`decimal` when a column's values can be parsed as either.

This addresses the long-standing feature request in issue #838. The existing
behaviour (inferring `decimal`) is preserved by default; set
`PreferFloats = true` to opt in to `float` inference.

Implementation:
- `StructuralInference.fs`: add `preferFloats: bool` parameter to
  `inferPrimitiveType`; guard the decimal match arm with `when not preferFloats`;
  add internal helper `getInferedTypeFromStringPreferFloats`.
- `CsvInference.fs`: thread `preferFloats` through `inferCellType`,
  `inferType`, `inferColumnTypes`, and `CsvFile.InferColumnTypes`.
- `HtmlInference.fs`: pass `false` for the new parameter (HTML provider
  is unaffected).
- `CsvProvider.fs`: register the new static parameter at index 19.
- `TypeProviderInstantiation.fs`: update `CsvProviderArgs` record and args
  array for the design-time test harness.
- `CsvProvider.fs` (tests): add two tests covering the new parameter.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

* ci: trigger CI checks

---------

Co-authored-by: Repo Assist <copilot@github.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: Don Syme <dsyme@users.noreply.github.com>

Author: 4 people

src/FSharp.Data.Csv.Core/CsvInference.fs

@@ -125,6 +125,7 @@ let internal inferCellType
     missingValues
     inferenceMode
     strictBooleans
+    preferFloats
     cultureInfo
     unit
     (value: string)
@@ -140,7 +141,15 @@ let internal inferCellType
         InferedType.Null
     else
         let inferedType =
-            StructuralInference.getInferedTypeFromString unitsOfMeasureProvider inferenceMode cultureInfo value unit
+            if preferFloats then
+                StructuralInference.getInferedTypeFromStringPreferFloats
+                    unitsOfMeasureProvider
+                    inferenceMode
+                    cultureInfo
+                    value
+                    unit
+            else
+                StructuralInference.getInferedTypeFromString unitsOfMeasureProvider inferenceMode cultureInfo value unit
 
         if strictBooleans then
             // With StrictBooleans=true, only "true"/"false" trigger bool inference.
@@ -305,6 +314,7 @@ let internal inferType
     missingValues
     inferenceMode
     strictBooleans
+    preferFloats
     cultureInfo
     assumeMissingValues
     preferOptionals
@@ -358,6 +368,7 @@ let internal inferType
                                     missingValues
                                     inferenceMode
                                     strictBooleans
+                                    preferFloats
                                     cultureInfo
                                     unit
                                     value
@@ -456,6 +467,7 @@ let internal inferColumnTypes
     missingValues
     inferenceMode
     strictBooleans
+    preferFloats
     cultureInfo
     assumeMissingValues
     preferOptionals
@@ -469,6 +481,7 @@ let internal inferColumnTypes
         missingValues
         inferenceMode
         strictBooleans
+        preferFloats
         cultureInfo
         assumeMissingValues
         preferOptionals
@@ -497,7 +510,8 @@ type CsvFile with
             assumeMissingValues,
             preferOptionals,
             unitsOfMeasureProvider,
-            ?strictBooleans
+            ?strictBooleans,
+            ?preferFloats
         ) =
 
         let headerNamesAndUnits, schema =
@@ -511,6 +525,7 @@ type CsvFile with
             missingValues
             inferenceMode
             (defaultArg strictBooleans false)
+            (defaultArg preferFloats false)
             cultureInfo
             assumeMissingValues
             preferOptionals

src/FSharp.Data.DesignTime/Csv/CsvProvider.fs

@@ -58,6 +58,7 @@ type public CsvProvider(cfg: TypeProviderConfig) as this =
         let preferDateOnly = args.[16] :?> bool
         let strictBooleans = args.[17] :?> bool
         let useOriginalNames = args.[18] :?> bool
+        let preferFloats = args.[19] :?> bool
 
         // This provider already has a schema mechanism, so let's disable inline schemas.
         let inferenceMode = InferenceMode'.ValuesOnly
@@ -116,7 +117,8 @@ type public CsvProvider(cfg: TypeProviderConfig) as this =
                         assumeMissingValues,
                         preferOptionals,
                         unitsOfMeasureProvider,
-                        strictBooleans
+                        strictBooleans,
+                        preferFloats
                     )
 #if NET6_0_OR_GREATER
                 if preferDateOnly && ProviderHelpers.runtimeSupportsNet6Types cfg.RuntimeAssembly then
@@ -241,7 +243,8 @@ type public CsvProvider(cfg: TypeProviderConfig) as this =
           ProvidedStaticParameter("EmbeddedResource", typeof<string>, parameterDefaultValue = "")
           ProvidedStaticParameter("PreferDateOnly", typeof<bool>, parameterDefaultValue = false)
           ProvidedStaticParameter("StrictBooleans", typeof<bool>, parameterDefaultValue = false)
-          ProvidedStaticParameter("UseOriginalNames", typeof<bool>, parameterDefaultValue = false) ]
+          ProvidedStaticParameter("UseOriginalNames", typeof<bool>, parameterDefaultValue = false)
+          ProvidedStaticParameter("PreferFloats", typeof<bool>, parameterDefaultValue = false) ]
 
     let helpText =
         """<summary>Typed representation of a CSV file.</summary>

src/FSharp.Data.Html.Core/HtmlInference.fs

@@ -28,6 +28,7 @@ let internal inferColumns parameters (headerNamesAndUnits: _[]) rows =
         parameters.MissingValues
         parameters.InferenceMode
         false
+        false
         parameters.CultureInfo
         assumeMissingValues
         parameters.PreferOptionals

src/FSharp.Data.Runtime.Utilities/StructuralInference.fs

@@ -525,6 +525,7 @@ let inferPrimitiveType
     (unitsOfMeasureProvider: IUnitsOfMeasureProvider)
     (inferenceMode: InferenceMode')
     (cultureInfo: CultureInfo)
+    (preferFloats: bool)
     (value: string)
     (desiredUnit: Type option)
     =
@@ -573,7 +574,7 @@ let inferPrimitiveType
             makePrimitive typeof<DateOnly>
 #endif
         | Parse TextConversions.AsDateTime date when not (isFakeDate date value) -> makePrimitive typeof<DateTime>
-        | Parse TextConversions.AsDecimal _ -> makePrimitive typeof<decimal>
+        | Parse TextConversions.AsDecimal _ when not preferFloats -> makePrimitive typeof<decimal>
         | Parse (TextConversions.AsFloat [||] false) _ -> makePrimitive typeof<float>
         | Parse asGuid _ -> makePrimitive typeof<Guid>
         | _ -> None
@@ -622,7 +623,11 @@ let inferPrimitiveType
 /// Infers the type of a simple string value
 [<Obsolete("This API will be made internal in a future release. Please file an issue at https://github.com/fsprojects/FSharp.Data/issues/1458 if you need this public.")>]
 let getInferedTypeFromString unitsOfMeasureProvider inferenceMode cultureInfo value unit =
-    inferPrimitiveType unitsOfMeasureProvider inferenceMode cultureInfo value unit
+    inferPrimitiveType unitsOfMeasureProvider inferenceMode cultureInfo false value unit
+
+/// Infers the type of a simple string value, preferring float over decimal
+let internal getInferedTypeFromStringPreferFloats unitsOfMeasureProvider inferenceMode cultureInfo value unit =
+    inferPrimitiveType unitsOfMeasureProvider inferenceMode cultureInfo true value unit
 
 #if NET6_0_OR_GREATER
 /// Replaces DateOnly → DateTime and TimeOnly → TimeSpan throughout an InferedType tree.

tests/FSharp.Data.DesignTime.Tests/InferenceTests.fs

@@ -24,7 +24,7 @@ let internal unitsOfMeasureProvider = ProviderHelpers.unitsOfMeasureProvider
 
 let internal inferType (csv:CsvFile) inferRows missingValues cultureInfo schema assumeMissingValues preferOptionals =
     let headerNamesAndUnits, schema = parseHeaders csv.Headers csv.NumberOfColumns schema unitsOfMeasureProvider
-    inferType headerNamesAndUnits schema (csv.Rows |> Seq.map (fun x -> x.Columns)) inferRows missingValues inferenceMode false cultureInfo assumeMissingValues preferOptionals unitsOfMeasureProvider
+    inferType headerNamesAndUnits schema (csv.Rows |> Seq.map (fun x -> x.Columns)) inferRows missingValues inferenceMode false false cultureInfo assumeMissingValues preferOptionals unitsOfMeasureProvider
 
 let internal toRecord fields = InferedType.Record(None, fields, false)
 
@@ -411,7 +411,7 @@ let ``Doesn't infer 12-002 as a date``() =
 
 [<Test>]
 let ``Doesn't infer ad3mar as a date``() =
-  StructuralInference.inferPrimitiveType unitsOfMeasureProvider inferenceMode CultureInfo.InvariantCulture "ad3mar" None
+  StructuralInference.inferPrimitiveType unitsOfMeasureProvider inferenceMode CultureInfo.InvariantCulture false "ad3mar" None
   |> should equal (InferedType.Primitive(typeof<string>, None, false, false))
 
 [<Test>]

tests/FSharp.Data.DesignTime.Tests/TypeProviderInstantiation.fs

@@ -28,7 +28,8 @@ type internal CsvProviderArgs =
       EmbeddedResource : string
       PreferDateOnly : bool
       StrictBooleans : bool
-      UseOriginalNames : bool }
+      UseOriginalNames : bool
+      PreferFloats : bool }
 
 type internal XmlProviderArgs =
     { Sample : string
@@ -106,7 +107,8 @@ type internal TypeProviderInstantiation =
                    box x.EmbeddedResource
                    box x.PreferDateOnly
                    box x.StrictBooleans
-                   box x.UseOriginalNames |]
+                   box x.UseOriginalNames
+                   box x.PreferFloats |]
             | Xml x ->
                 (fun cfg -> new XmlProvider(cfg) :> TypeProviderForNamespaces),
                 [| box x.Sample
@@ -246,7 +248,8 @@ type internal TypeProviderInstantiation =
                   EmbeddedResource = ""
                   PreferDateOnly = false
                   StrictBooleans = false
-                  UseOriginalNames = false }
+                  UseOriginalNames = false
+                  PreferFloats = false }
         | "Xml" ->
             Xml { Sample = args.[1]
                   SampleIsList = args.[2] |> bool.Parse

tests/FSharp.Data.Tests/CsvProvider.fs

@@ -766,3 +766,21 @@ let ``CsvProvider Row With* methods do not mutate the original row`` () =
     let row = csv.Rows |> Seq.head
     let _ = row.WithName("Charlie")
     row.Name |> should equal "Alice"
+
+// Tests for PreferFloats static parameter (issue #838)
+let [<Literal>] csvWithDecimals = "Name,Price,Rate\nAlice,9.99,1.5\nBob,12.50,2.7"
+
+type CsvPreferFloats = CsvProvider<csvWithDecimals, PreferFloats = true>
+type CsvDefaultDecimal = CsvProvider<csvWithDecimals>
+
+[<Test>]
+let ``CsvProvider PreferFloats=true infers float instead of decimal`` () =
+    let row = CsvPreferFloats.GetSample().Rows |> Seq.head
+    row.Price |> should equal 9.99
+    row.Rate |> should equal 1.5
+
+[<Test>]
+let ``CsvProvider PreferFloats=false (default) infers decimal for decimal values`` () =
+    let row = CsvDefaultDecimal.GetSample().Rows |> Seq.head
+    row.Price |> should equal 9.99m
+    row.Rate |> should equal 1.5m