refactor: convert to a Cargo workspace

CLAUDE.md

@@ -26,15 +26,15 @@ cargo clippy             # lint
 cargo fmt --check        # check formatting
 ```
 
-Shell-based equivalence and integration tests live in `tests/`:
-- `tests/equivalence-e2e.sh` — end-to-end comparison against original MaxBin2
-- `tests/pipeline-trace.sh` — recursive pipeline equivalence (Rust vs C++)
-- `tests/cli-equivalence.sh` / `tests/cli-integration.sh` — CLI behavior
-- `tests/bench-genecaller.sh` — FragGeneScan vs FragGeneScanRs benchmark
+Shell-based equivalence and integration tests live in `crates/maxbin-rs/tests/`:
+- `crates/maxbin-rs/tests/equivalence-e2e.sh` — end-to-end comparison against original MaxBin2
+- `crates/maxbin-rs/tests/pipeline-trace.sh` — recursive pipeline equivalence (Rust vs C++)
+- `crates/maxbin-rs/tests/cli-equivalence.sh` / `crates/maxbin-rs/tests/cli-integration.sh` — CLI behavior
+- `crates/maxbin-rs/tests/bench-genecaller.sh` — FragGeneScan vs FragGeneScanRs benchmark
 
 ## Architecture
 
-### CLI (clap derive, `src/cli.rs`)
+### CLI (clap derive, `crates/maxbin-rs/src/cli.rs`)
 
 Uses standard double-dash flags (`--contig`, `--abund`, etc.). Single-dash compatibility was dropped in v0.2.0. Subcommands expose individual pipeline stages:
 
@@ -45,7 +45,7 @@ Uses standard double-dash flags (`--contig`, `--abund`, etc.). Single-dash compa
 - `maxbin-rs cpp-em` — run the original C++ EM via FFI (equivalence testing only)
 - `maxbin-rs sam-to-abund` — compute abundance from a SAM file
 
-### Key source files
+### Key source files (in `crates/maxbin-rs/src/`)
 
 - `pipeline.rs` — orchestration for each subcommand
 - `emanager.rs` — EM algorithm core (E-step, M-step, convergence)

Cargo.toml

@@ -1,26 +1,10 @@
-[package]
-name = "maxbin-rs"
-version = "0.4.0"
-edition = "2024"
+[workspace]
+resolver = "3"
+members = ["crates/maxbin-rs"]
 
-[dependencies]
-clap = { version = "4.6.1", features = ["derive"] }
-flate2 = "1.1.9"
-frag_gene_scan_rs = "1.1.0"
-rayon = "1.10"
-serde = { version = "1", features = ["derive"] }
-serde_json = "1"
-seq_io = "0.3"
-tempfile = "3"
-
-[build-dependencies]
-cc = "1.2.60"
-
-[dev-dependencies]
-proptest = "1.11.0"
-
-# Optimized test profile for benchmarks.
+# Profiles are only honored at the workspace root, so the benchmark profile
+# that used to live in the crate manifest moves here.
 # Usage: cargo nextest run --cargo-profile bench bench_components -- --ignored
 [profile.bench]
 inherits = "release"
-debug = true  # keep symbols for profiling
+debug = true # keep symbols for profiling

crates/maxbin-rs/Cargo.toml

@@ -0,0 +1,20 @@
+[package]
+name = "maxbin-rs"
+version = "0.4.0"
+edition = "2024"
+
+[dependencies]
+clap = { version = "4.6.1", features = ["derive"] }
+flate2 = "1.1.9"
+frag_gene_scan_rs = "1.1.0"
+rayon = "1.10"
+serde = { version = "1", features = ["derive"] }
+serde_json = "1"
+seq_io = "0.3"
+tempfile = "3"
+
+[build-dependencies]
+cc = "1.2.60"
+
+[dev-dependencies]
+proptest = "1.11.0"

build.rs → crates/maxbin-rs/build.rs

@@ -50,7 +50,7 @@ fn main() {
     let manifest_dir = PathBuf::from(env::var("CARGO_MANIFEST_DIR").unwrap());
 
     // Rebuild if patch or FFI wrappers change.
-    println!("cargo:rerun-if-changed=nix/maxbin2-cpp-ffi.patch");
+    println!("cargo:rerun-if-changed=vendor/maxbin2-cpp-ffi.patch");
     println!("cargo:rerun-if-changed=vendor/ffi");
     println!("cargo:rerun-if-env-changed=MAXBIN2_SRC_TARBALL");
 
@@ -108,7 +108,7 @@ fn main() {
     fs::remove_dir(&src_subdir).unwrap();
 
     // Step 3: Apply the patch.
-    let patch_path = manifest_dir.join("nix/maxbin2-cpp-ffi.patch");
+    let patch_path = manifest_dir.join("vendor/maxbin2-cpp-ffi.patch");
     // The patch has paths like a/src/EManager.cpp, so we need to adjust.
     // We'll apply with -p2 to strip "a/src/" down to the filename.
     let patch_output = Command::new("patch")

flake.nix

@@ -536,7 +536,7 @@
               ''
                 export MAXBIN_RS_DETERMINISTIC=1
                 export HOME=$(mktemp -d)
-                bash ${./tests/pipeline-trace.sh} \
+                bash ${./crates/maxbin-rs/tests/pipeline-trace.sh} \
                   "${intermediates.cami-small}/contigs.fa" \
                   "${intermediates.cami-small}/abund" \
                   "${intermediates.cami-small}/hmmout"

nix/tests.nix

@@ -2,7 +2,7 @@
 #
 # Each entry here creates a small shell script that:
 #   1. Sets environment variables pointing to Nix-cached test data
-#   2. Runs the actual test script from tests/
+#   2. Runs the actual test script from crates/maxbin-rs/tests/
 #
 # This indirection exists because the test scripts are plain bash (readable
 # without knowing Nix), but they need paths to data in the Nix store —
@@ -31,7 +31,7 @@
 
 let
   # Helper: create a sandboxed CLI integration test for a given dataset.
-  # Runs tests/cli-integration.sh with the binary on PATH and test data
+  # Runs crates/maxbin-rs/tests/cli-integration.sh with the binary on PATH and test data
   # available. Tests flag parsing, subcommands, error cases, and output
   # validation end-to-end.
   mkCliTest =
@@ -55,7 +55,7 @@ let
         export MAXBIN_RS_DETERMINISTIC=1
         export HOME=$(mktemp -d)
 
-        bash ${../tests/cli-integration.sh}
+        bash ${../crates/maxbin-rs/tests/cli-integration.sh}
 
         touch $out
       '';
@@ -83,7 +83,7 @@ let
         export MAXBIN_RS_DETERMINISTIC=1
         export HOME=$(mktemp -d)
 
-        bash ${../tests/cli-equivalence.sh}
+        bash ${../crates/maxbin-rs/tests/cli-equivalence.sh}
 
         touch $out
       '';
@@ -114,7 +114,7 @@ let
         maxbin-rs
       ];
       text = ''
-        bash ${../tests/pipeline-trace.sh} \
+        bash ${../crates/maxbin-rs/tests/pipeline-trace.sh} \
           "${contigs}" \
           "${intermediates'}/abund" \
           "${intermediates'}/hmmout"
@@ -443,7 +443,7 @@ in
       export CONTIGS="${datasets.bfragilis.contigs}"
       export TRAIN_DIR="${fraggenescan-rs}/share/FragGeneScanRs/train"
       export PATH="${fraggenescan}/libexec/FragGeneScan:$PATH"
-      bash ${../tests/bench-genecaller.sh}
+      bash ${../crates/maxbin-rs/tests/bench-genecaller.sh}
     '';
   };
 
@@ -456,7 +456,7 @@ in
       export CONTIGS="${datasets.capes-s7.contigs}"
       export TRAIN_DIR="${fraggenescan-rs}/share/FragGeneScanRs/train"
       export PATH="${fraggenescan}/libexec/FragGeneScan:$PATH"
-      bash ${../tests/bench-genecaller.sh}
+      bash ${../crates/maxbin-rs/tests/bench-genecaller.sh}
     '';
   };
 
@@ -469,7 +469,7 @@ in
     text = ''
       export CONTIGS="${datasets.bfragilis.contigs}"
       export ABUND="${intermediates.bfragilis}/abund"
-      bash ${../tests/bench-pipeline.sh}
+      bash ${../crates/maxbin-rs/tests/bench-pipeline.sh}
     '';
   };
 
@@ -479,7 +479,7 @@ in
     text = ''
       export CONTIGS="${datasets.capes-s7.contigs}"
       export ABUND="${intermediates.capes}/abund"
-      bash ${../tests/bench-pipeline.sh}
+      bash ${../crates/maxbin-rs/tests/bench-pipeline.sh}
     '';
   };
 
@@ -490,7 +490,7 @@ in
     text = ''
       export CONTIGS="${datasets.capes-s7.contigs}"
       export READS="${datasets.capes-s7.reads1} ${datasets.capes-s7.reads2}"
-      bash ${../tests/bench-pipeline.sh}
+      bash ${../crates/maxbin-rs/tests/bench-pipeline.sh}
     '';
   };