feat: compare array metadata against checked in goldenfiles (#1812)

This is part of a general epic (is that word taboo?) around stability. I
have a separate changeset that stores an entire Vortex file and checks
for round-trip compatibility.

However, it's also helpful to know on a per-encoding level that the
metadata format is stable.

This PR adds [goldenfile](https://crates.io/crates/goldenfile) tests for
all encoding metadata. This is a bit simplistic: we don't check for
backward/forward compatibility, this is just an attempt to make it
clearer on PRs when changes are made to the metadata format.

Now when changes are made to Array metadata, you should re-run the test
with

```
UPDATE_GOLDENFILES=1 cargo test -p ...
```

and that will update the checked-in metadatas.



Roughly half of the changed files are checked-in snapshots, most of the
other half are simple test definitions. The only serious change is the
addition of the `check_metadata` helper in test-utils.

Author: a10y

Cargo.lock

@@ -85,6 +85,7 @@ futures = { version = "0.3", default-features = false }
 futures-executor = "0.3"
 futures-util = "0.3"
 getrandom = "0.2.14"
+goldenfile = "1"
 half = { version = "^2", features = ["std", "num-traits"] }
 hashbrown = "0.15.0"
 homedir = "0.3.3"

Cargo.toml

@@ -28,9 +28,9 @@ vortex-fastlanes = { workspace = true }
 vortex-scalar = { workspace = true }
 
 [dev-dependencies]
-arrow-array = { workspace = true }
 divan = { workspace = true }
 rstest = { workspace = true }
+vortex-array = { path = "../../vortex-array", features = ["test_util"] }
 
 [[bench]]
 name = "alp_compress"

encodings/alp/Cargo.toml

@@ -21,8 +21,8 @@ impl_encoding!("vortex.alp", ids::ALP, ALP);
 
 #[derive(Debug, Clone, Serialize, Deserialize)]
 pub struct ALPMetadata {
-    exponents: Exponents,
-    patches: Option<PatchesMetadata>,
+    pub(crate) exponents: Exponents,
+    pub(crate) patches: Option<PatchesMetadata>,
 }
 
 impl Display for ALPMetadata {
@@ -150,3 +150,27 @@ impl VisitorVTable<ALPArray> for ALPEncoding {
 }
 
 impl StatisticsVTable<ALPArray> for ALPEncoding {}
+
+#[cfg(test)]
+mod tests {
+    use vortex_array::patches::PatchesMetadata;
+    use vortex_array::test_utils::check_metadata;
+    use vortex_dtype::PType;
+
+    use crate::{ALPMetadata, Exponents};
+
+    #[cfg_attr(miri, ignore)]
+    #[test]
+    fn test_alp_metadata() {
+        check_metadata(
+            "alp.metadata",
+            ALPMetadata {
+                patches: Some(PatchesMetadata::new(usize::MAX, PType::U64)),
+                exponents: Exponents {
+                    e: u8::MAX,
+                    f: u8::MAX,
+                },
+            },
+        );
+    }
+}

encodings/alp/goldenfiles/alp.metadata

@@ -258,9 +258,27 @@ impl ArrayTrait for ALPRDArray {}
 mod test {
     use rstest::rstest;
     use vortex_array::array::PrimitiveArray;
+    use vortex_array::patches::PatchesMetadata;
+    use vortex_array::test_utils::check_metadata;
     use vortex_array::{IntoArrayData, IntoCanonical};
+    use vortex_dtype::PType;
 
-    use crate::{alp_rd, ALPRDFloat};
+    use crate::{alp_rd, ALPRDFloat, ALPRDMetadata};
+
+    #[cfg_attr(miri, ignore)]
+    #[test]
+    fn test_alprd_metadata() {
+        check_metadata(
+            "alprd.metadata",
+            ALPRDMetadata {
+                right_bit_width: u8::MAX,
+                patches: Some(PatchesMetadata::new(usize::MAX, PType::U64)),
+                dict: [0u16; 8],
+                left_parts_ptype: PType::U64,
+                dict_len: 8,
+            },
+        );
+    }
 
     #[rstest]
     #[case(vec![0.1f32.next_up(); 1024], 1.123_848_f32)]

encodings/alp/goldenfiles/alprd.metadata

@@ -25,3 +25,6 @@ vortex-buffer = { workspace = true }
 vortex-dtype = { workspace = true }
 vortex-error = { workspace = true }
 vortex-scalar = { workspace = true }
+
+[dev-dependencies]
+vortex-array = { path = "../../vortex-array", features = ["test_util"] }

encodings/alp/src/alp/array.rs

@@ -133,10 +133,22 @@ impl VisitorVTable<ByteBoolArray> for ByteBoolEncoding {
 
 #[cfg(test)]
 mod tests {
+    use vortex_array::test_utils::check_metadata;
     use vortex_array::validity::ArrayValidity;
 
     use super::*;
 
+    #[cfg_attr(miri, ignore)]
+    #[test]
+    fn test_bytebool_metadata() {
+        check_metadata(
+            "bytebool.metadata",
+            ByteBoolMetadata {
+                validity: ValidityMetadata::AllValid,
+            },
+        );
+    }
+
     #[test]
     fn test_validity_construction() {
         let v = vec![true, false];