Zig drop-in replacement for bcrham and ig-sw (zero C++ deps)

.gitmodules

@@ -25,3 +25,6 @@
 [submodule "packages/ig-sw"]
 	path = packages/ig-sw
 	url = https://github.com/matsengrp/ig-sw.git
+[submodule "zig-backend/partis-zig-core"]
+	path = zig-backend/partis-zig-core
+	url = https://github.com/matsengrp/partis-zig-core.git

bin/install-zig-backend.sh

@@ -0,0 +1,117 @@
+#!/usr/bin/env bash
+# Install the partis-zig-core Zig backend (drop-in replacements for bcrham and ig-sw).
+#
+# Usage:
+#   bin/install-zig-backend.sh [--install-dir DIR]
+#
+# After running, pass these flags to partis:
+#   --bcrham-binary <install-dir>/partis-zig-core/zig-out/bin/partis-zig-core
+#   --ig-sw-binary  <install-dir>/partis-zig-core/zig-out/bin/partis-zig-igsw
+#
+# The Zig backend requires no external libraries (GSL, OpenBLAS, etc.) and
+# produces bit-for-bit identical annotations to the C++ bcrham and C ig-sw binaries.
+set -euo pipefail
+
+ZIG_VERSION="0.15.2"
+ZIG_REPO="https://github.com/matsengrp/partis-zig-core.git"
+DEFAULT_INSTALL_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)/zig-backend"
+
+INSTALL_DIR="$DEFAULT_INSTALL_DIR"
+while [[ $# -gt 0 ]]; do
+    case "$1" in
+        --install-dir) INSTALL_DIR="$2"; shift 2 ;;
+        *) echo "Unknown argument: $1" >&2; exit 1 ;;
+    esac
+done
+
+ZIG_CORE_DIR="$INSTALL_DIR/partis-zig-core"
+ZIG_EXE_FINAL="$ZIG_CORE_DIR/zig-out/bin/partis-zig-core"
+
+echo "==> Installing partis Zig backend to: $INSTALL_DIR"
+mkdir -p "$INSTALL_DIR"
+
+# ── 1. Ensure Zig is available ──────────────────────────────────────────────
+
+find_zig() {
+    # Prefer a zig in PATH that matches the required version
+    if command -v zig &>/dev/null; then
+        local v
+        v=$(zig version 2>/dev/null || echo "")
+        if [[ "$v" == "$ZIG_VERSION" ]]; then
+            echo "zig"
+            return
+        fi
+    fi
+    # Check if we already downloaded it
+    local downloaded="$INSTALL_DIR/zig-$ZIG_VERSION/zig"
+    if [[ -x "$downloaded" ]]; then
+        echo "$downloaded"
+        return
+    fi
+    echo ""
+}
+
+ZIG=$(find_zig)
+if [[ -z "$ZIG" ]]; then
+    echo "==> Downloading Zig $ZIG_VERSION..."
+    OS=$(uname -s | tr '[:upper:]' '[:lower:]')
+    ARCH=$(uname -m)
+    # Normalize arch names
+    case "$ARCH" in
+        x86_64)  ARCH="x86_64" ;;
+        aarch64|arm64) ARCH="aarch64" ;;
+        *) echo "ERROR: Unsupported architecture: $ARCH" >&2; exit 1 ;;
+    esac
+    case "$OS" in
+        linux)  OS_TAG="linux" ;;
+        darwin) OS_TAG="macos" ;;
+        *) echo "ERROR: Unsupported OS: $OS" >&2; exit 1 ;;
+    esac
+    ZIG_TARBALL="zig-${OS_TAG}-${ARCH}-${ZIG_VERSION}.tar.xz"
+    ZIG_URL="https://ziglang.org/download/${ZIG_VERSION}/${ZIG_TARBALL}"
+    ZIG_DIR="$INSTALL_DIR/zig-${ZIG_VERSION}"
+    mkdir -p "$ZIG_DIR"
+    curl -fsSL "$ZIG_URL" | tar -xJ -C "$ZIG_DIR" --strip-components=1
+    ZIG="$ZIG_DIR/zig"
+    echo "    Zig $ZIG_VERSION installed at: $ZIG"
+else
+    echo "==> Using Zig: $ZIG ($(zig version 2>/dev/null || $ZIG version))"
+fi
+
+# ── 2. Clone or update partis-zig-core ──────────────────────────────────────
+
+if [[ -d "$ZIG_CORE_DIR/.git" ]]; then
+    echo "==> Updating partis-zig-core..."
+    git -C "$ZIG_CORE_DIR" pull --ff-only
+else
+    echo "==> Cloning partis-zig-core..."
+    git clone "$ZIG_REPO" "$ZIG_CORE_DIR"
+fi
+
+# ── 3. Build ─────────────────────────────────────────────────────────────────
+
+# build.zig references ig-sw sources inside the partis repo.
+# Set PARTIS_DIR so build.zig can find them.
+PARTIS_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+export PARTIS_DIR
+
+echo "==> Building partis-zig-core (this takes ~30s)..."
+(cd "$ZIG_CORE_DIR" && "$ZIG" build -Doptimize=ReleaseFast)
+
+if [[ ! -x "$ZIG_EXE_FINAL" ]]; then
+    echo "ERROR: build succeeded but executable not found at $ZIG_EXE_FINAL" >&2
+    exit 1
+fi
+
+ZIG_IGSW_EXE_FINAL="$ZIG_CORE_DIR/zig-out/bin/partis-zig-igsw"
+
+echo ""
+echo "Build complete:"
+echo "  $ZIG_EXE_FINAL"
+echo "  $ZIG_IGSW_EXE_FINAL"
+echo ""
+echo "To use the Zig backend, pass these flags to partis:"
+echo "  --bcrham-binary $ZIG_EXE_FINAL"
+echo "  --ig-sw-binary $ZIG_IGSW_EXE_FINAL"
+echo ""
+echo "Omit these flags to use the default C++ bcrham and C ig-sw binaries."

bin/partis

@@ -1489,6 +1489,7 @@ parent_args.append({'name' : '--count-correlations', 'kwargs' : {'action' : 'sto
 parent_args.append({'name' : '--dont-write-parameters', 'kwargs' : {'action' : 'store_true', 'help' : 'don\'t write parameters to disk even if you\'ve counted them (mostly for use in conjunction with --only-smith-waterman, so you can avoid cluttering up your file system)'}})
 parent_args.append({'name' : '--partis-dir', 'kwargs' : {'default' : partis_dir, 'help' : 'for internal use only'}})
 parent_args.append({'name' : '--ig-sw-binary', 'kwargs' : {'default' : partis_dir + '/packages/ig-sw/src/ig_align/ig-sw', 'help' : 'Path to ig-sw executable.'}})
+parent_args.append({'name' : '--bcrham-binary', 'kwargs' : {'default' : partis_dir + '/packages/ham/bcrham', 'help' : 'Path to bcrham executable.'}})
 parent_args.append({'name' : '--vsearch-binary', 'kwargs' : { 'help' : 'Path to vsearch binary (vsearch binaries for linux and darwin are pre-installed in bin/, but for other systems you need to get your own)'}})
 parent_args.append({'name' : '--is-simu', 'kwargs' : {'action' : 'store_true', 'help' : 'Set if running on simulated sequences'}})
 parent_args.append({'name' : '--skip-unproductive', 'kwargs' : {'action' : 'store_true', 'help' : 'Skip sequences which Smith-Waterman determines to be unproductive (i.e. if they have stop codons, out of frame cdr3, or mutated cyst/tryp/phen)'}})

partis/partitiondriver.py

@@ -809,8 +809,8 @@ def get_n_calculated_per_process(self):
 
         total = 0.  # sum over each process
         for procinfo in self.bcrham_proc_info:
-            if 'vtb' not in procinfo['calcd'] or 'fwd' not in procinfo['calcd']:
-                print('%s couldn\'t find vtb/fwd in:\n%s' % (utils.color('red', 'warning'), procinfo['calcd']))  # may as well not fail, it probably just means we lost some stdout somewhere. Which, ok, is bad, but let's say it shouldn't be fatal.
+            if procinfo['calcd'].get('vtb') is None or procinfo['calcd'].get('fwd') is None:
+                print('%s couldn\'t find vtb/fwd in:\n%s' % (utils.color('red', 'warning'), procinfo['calcd']))  # may as well not fail, it probably just means we lost some stdout somewhere (or are using the Zig backend which doesn't emit calcd debug strings). Which, ok, is bad, but let's say it shouldn't be fatal.
                 return 1.  # er, or something?
             if self.args.naive_hamming_cluster:  # make sure we didn't accidentally calculate some fwds
                 assert procinfo['calcd']['fwd'] == 0.
@@ -1263,7 +1263,7 @@ def dbgstr():
     # ----------------------------------------------------------------------------------------
     def get_hmm_cmd_str(self, algorithm, csv_infname, csv_outfname, parameter_dir, precache_all_naive_seqs, n_procs):
         """ Return the appropriate bcrham command string """
-        cmd_str = self.args.partis_dir + '/packages/ham/bcrham'
+        cmd_str = self.args.bcrham_binary
         cmd_str += ' --algorithm ' + algorithm
         if self.args.debug > 0:
             cmd_str += ' --debug ' + str(self.args.debug)

test/test.py

@@ -424,6 +424,10 @@ def run(self, args):
                 if tkey in info:
                     cmd_str += ' --%s %s' % (tkey if tkey != 'inpath' else self.astr('in', dt=idt), info[tkey])
             cmd_str += ' %s' % ' '.join(info['extras'] + self.common_extras)
+            if args.bcrham_binary is not None:
+                cmd_str += ' --bcrham-binary %s' % args.bcrham_binary
+            if args.ig_sw_binary is not None:
+                cmd_str += ' --ig-sw-binary %s' % args.ig_sw_binary
 
             if name == 'simulate':
                 cmd_str += ' --%s %s' % (self.astr('out'), self.inpath('new', 'simu'))
@@ -941,6 +945,8 @@ def compare_files(fname):
 
 parser.add_argument('--glfo-dir', default='data/germlines/human')
 parser.add_argument('--locus', default='igh')
+parser.add_argument('--bcrham-binary', default=None, help='override default bcrham binary path (e.g. Zig drop-in replacement)')
+parser.add_argument('--ig-sw-binary', default=None, help='override default ig-sw binary path (e.g. Zig drop-in replacement)')
 args = parser.parse_args()
 assert not (args.quick and args.slow)  # it just doesn't make sense
 assert not (args.quick and args.paired)  # --quick ignores --paired, which is confusing