Support nested catListTable (by represented nested arrays as text) (#…

…242)

This is one possible "fix" to #168. With this we can `catListTable` arbitrarily deep trees of `ListTable`s.

It comes at a relatively high cost, however.

Currently we represent nested arrays with anonymous records. This works reasonably well, except that we can't extract the field from the anonymous record when we need it (PostgreSQL [theoretically](https://www.postgresql.org/docs/13/release-13.html#id-1.11.6.16.5.6) suports `.f1` syntax since PG13 but it only works in very limited situations). But it does mean we can decode the results using Hasql's binary decoders, and ordering works how we expect ('array[row(array[9])] < array[row(array[10])]'.

What this PR does is instead represent nested arrays as text. To be able to decode this, we need each 'DBType' to supply a text parser in addition to a binary decoder. It also means that ordering is no longer intuitive, because `array[array[9]::text] > array[array[10]::text]`. However, it does mean we can nest `catListTable`s to our heart's content and it will always just work.

changelog.d/20230707_185221_ollie_scriv.md

@@ -1,3 +1,7 @@
 ### Fixed
 
-- A partial fix for nesting `catListTable`. The underlying issue in [#168](https://github.com/circuithub/rel8/issues/168) is still present, but we've made a bit of progress to reduce when this bug can happen. ([#243](https://github.com/circuithub/rel8/pull/243))
+- A fix for [#168](https://github.com/circuithub/rel8/issues/168), which prevented using `catListTable` on arrays of arrays. To achieve this we had to coerce arrays of arrays to text internally, which unfortunately isn't completely transparent; you can oberve it if you write something like `listTable [listTable [10]] > listTable [listTable [9]]`: previously that would be `false`, but now it's `true`. Arrays of non-arrays are unaffected by this.
+
+### Changed
+
+- `TypeInformation`'s `decoder` field has changed. Instead of taking a `Hasql.Decoder`, it now takes a `Rel8.Decoder`, which itself is comprised of a `Hasql.Decoder` and an `attoparsec` `Parser`. This is necessitated by the fix for [#168](https://github.com/circuithub/rel8/issues/168); we generally decode things in PostgreSQL's binary format (using a `Hasql.Decoder`), but for nested arrays we now get things in PostgreSQL's text format (for which we need an `attoparsec` `Parser`), so must have both. Most `DBType` instances that use `mapTypeInformation` or `ParseTypeInformation`, or `DerivingVia` helpers like `ReadShow`, `JSONBEncoded`, `Enum` and `Composite` are unaffected by this change.

rel8.cabal

@@ -20,13 +20,16 @@ source-repository head
 library
   build-depends:
       aeson
+    , attoparsec
     , base ^>= 4.14 || ^>=4.15 || ^>=4.16 || ^>=4.17
+    , base16 >= 1.0
     , base-compat ^>= 0.11 || ^>= 0.12 || ^>= 0.13
     , bifunctors
     , bytestring
     , case-insensitive
     , comonad
     , contravariant
+    , data-textual
     , hasql ^>= 1.6.1.2
     , network-ip
     , opaleye ^>= 0.10.0.0
@@ -40,8 +43,10 @@ library
     , these
     , time
     , transformers
+    , utf8-string
     , uuid
     , vector
+
   default-language:
     Haskell2010
   ghc-options:
@@ -201,6 +206,7 @@ library
     Rel8.Type
     Rel8.Type.Array
     Rel8.Type.Composite
+    Rel8.Type.Decoder
     Rel8.Type.Eq
     Rel8.Type.Enum
     Rel8.Type.Information
@@ -210,6 +216,9 @@ library
     Rel8.Type.Name
     Rel8.Type.Num
     Rel8.Type.Ord
+    Rel8.Type.Parser
+    Rel8.Type.Parser.ByteString
+    Rel8.Type.Parser.Time
     Rel8.Type.ReadShow
     Rel8.Type.Semigroup
     Rel8.Type.String

src/Rel8/Expr/Serialize.hs

@@ -23,6 +23,7 @@ import {-# SOURCE #-} Rel8.Expr ( Expr( Expr ) )
 import Rel8.Expr.Opaleye ( scastExpr )
 import Rel8.Schema.Null ( Unnullify, Nullity( Null, NotNull ), Sql, nullable )
 import Rel8.Type ( DBType, typeInformation )
+import Rel8.Type.Decoder (Decoder (..))
 import Rel8.Type.Information ( TypeInformation(..) )
 
 
@@ -44,6 +45,6 @@ slitExpr nullity info@TypeInformation {encode} =
 
 
 sparseValue :: Nullity a -> TypeInformation (Unnullify a) -> Hasql.Row a
-sparseValue nullity TypeInformation {decode} = case nullity of
-  Null -> Hasql.column $ Hasql.nullable decode
-  NotNull -> Hasql.column $ Hasql.nonNullable decode
+sparseValue nullity TypeInformation {decode = Decoder {binary}} = case nullity of
+  Null -> Hasql.column $ Hasql.nullable binary
+  NotNull -> Hasql.column $ Hasql.nonNullable binary