Reduce impact of datafusion features on the dependency tree (#1235)

In service of my one true passion (removing crate features), an attemtp
at reducing datafusion to use the smallest possible subset of features,
preferring to use smaller crates when possible.

Cargo.toml

@@ -48,7 +48,7 @@ allocator-api2 = "0.2.16"
 anyhow = "1.0"
 arbitrary = "1.3.2"
 arrayref = "0.3.7"
-arrow = { version = "53.0.0", features = ["prettyprint"] }
+arrow = { version = "53.0.0" }
 arrow-arith = "53.0.0"
 arrow-array = "53.0.0"
 arrow-buffer = "53.0.0"
@@ -69,7 +69,7 @@ clap = "4.5.13"
 criterion = { version = "0.5.1", features = ["html_reports"] }
 croaring = "2.1.0"
 csv = "1.3.0"
-datafusion = "42.0.0"
+datafusion = { version = "42.0.0", default-features = false }
 datafusion-common = "42.0.0"
 datafusion-execution = "42.0.0"
 datafusion-expr = "42.0.0"
@@ -208,9 +208,9 @@ use_debug = "deny"
 
 [profile.release]
 codegen-units = 1
-lto = "thin" # attempts to perform optimizations across all crates within the dependency graph
+lto = "thin"      # attempts to perform optimizations across all crates within the dependency graph
 
 [profile.bench]
 codegen-units = 16 # default for "release", which "bench" inherits
-lto = false # default
+lto = false        # default
 debug = true

bench-vortex/Cargo.toml

@@ -30,7 +30,13 @@ bytes = { workspace = true }
 bzip2 = { workspace = true }
 clap = { workspace = true, features = ["derive"] }
 csv = { workspace = true }
-datafusion = { workspace = true }
+datafusion = { workspace = true, features = [
+    "parquet",
+    "datetime_expressions",
+] }
+datafusion-common = { workspace = true }
+datafusion-physical-plan = { workspace = true }
+datafusion-execution = { workspace = true }
 enum-iterator = { workspace = true }
 flexbuffers = { workspace = true }
 futures = { workspace = true, features = ["executor"] }
@@ -41,7 +47,7 @@ itertools = { workspace = true }
 log = { workspace = true }
 mimalloc = { workspace = true }
 object_store = { workspace = true, features = ["aws"] }
-parquet = { workspace = true, features = [] }
+parquet = { workspace = true, features = ["async"] }
 prettytable-rs = { workspace = true }
 rand = { workspace = true }
 rayon = { workspace = true }

bench-vortex/benches/datafusion.rs

@@ -5,12 +5,12 @@ use arrow_array::RecordBatch;
 use arrow_schema::{DataType, Field, Schema};
 use criterion::measurement::Measurement;
 use criterion::{black_box, criterion_group, criterion_main, BenchmarkGroup, Criterion};
-use datafusion::common::Result as DFResult;
 use datafusion::datasource::{MemTable, TableProvider};
 use datafusion::execution::memory_pool::human_readable_size;
 use datafusion::functions_aggregate::count::count_distinct;
 use datafusion::logical_expr::lit;
 use datafusion::prelude::{col, DataFrame, SessionContext};
+use datafusion_common::Result as DFResult;
 use vortex::aliases::hash_set::HashSet;
 use vortex::compress::CompressionStrategy;
 use vortex::dict::DictEncoding;

bench-vortex/src/tpch/execute.rs

@@ -1,7 +1,7 @@
 use arrow_array::RecordBatch;
-use datafusion::common::Result;
-use datafusion::physical_plan::collect;
 use datafusion::prelude::SessionContext;
+use datafusion_common::Result;
+use datafusion_physical_plan::collect;
 
 use crate::tpch::Format;