Add support for the Tcl programming language

Based on PR #187. Creates a language-specific Tcl.dockerfile instead of
modifying the main Dockerfile. Tested with claude-haiku-4-5 on HumanEval
(1 completion, temperature 0.2): pass@1 = 84.5%.

Co-Authored-By: Rudolf Adamkovic <rudolf@adamkovic.org>
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: 2 people

dataset_builder/humaneval_to_tcl.py

@@ -0,0 +1,68 @@
+# This script translates problems from the OpenAI HumanEval dataset into Tcl.
+import re
+import ast
+from typing import List
+
+
+class Translator:
+
+    stop = ["\nproc ", "\n#", "\n\n"]
+
+    def file_ext(self):
+        return "tcl"
+
+    def translate_prompt(self, name: str, args: List[ast.arg], _returns, description: str):
+        tcl_description = (
+            "# " + re.sub(r"\n(\s*)", "\n# ", description.strip()) + "\n" if description else ""
+        )
+        arg_names = [arg.arg for arg in args]
+        arg_list = " ".join(arg_names)
+        return f"{tcl_description}proc {name} {{{arg_list}}} {{\n"
+
+    def test_suite_prefix_lines(self, entry_point: str) -> List[str]:
+        return [
+            "",
+            "proc assert_equal {actual expected} {",
+            "    if {$actual ne $expected} {",
+            "        error \"Expected: $expected | Actual: $actual\"",
+            "    }",
+            "}",
+            "",
+            f"interp alias {{}} candidate {{}} {entry_point}",
+            "",
+        ]
+
+    def test_suite_suffix_lines(self) -> List[str]:
+        return []
+
+    def deep_equality(self, left: str, right: str) -> str:
+        return f"assert_equal {left} {right}"
+
+    def gen_literal(self, c: bool | str | int | float | None):
+        if type(c) == bool:
+            return str(c).lower()
+        elif type(c) == str:
+            escaped = c.replace("\\", "\\\\").replace('"', '\\"').replace("$", "\\$").replace("[", "\\[").replace("\n", "\\n")
+            return f'"{escaped}"'
+        elif c is None:
+            return '""'
+        return repr(c)
+
+    def gen_var(self, v: str):
+        return v
+
+    def gen_list(self, l: List[str]):
+        return "[list " + " ".join(l) + "]"
+
+    def gen_tuple(self, t: List[str]):
+        return "[list " + " ".join(t) + "]"
+
+    def gen_dict(self, keys: List[str], values: List[str]):
+        pairs = " ".join(f"{k} {v}" for k, v in zip(keys, values))
+        return "[dict create " + pairs + "]"
+
+    def gen_call(self, func: str, args: List[str]):
+        return "[" + func + " " + " ".join(args) + "]"
+
+    def no_completion_prompt_stub(self):
+        return "    return 0\n}\n"

dataset_builder/terms.csv

@@ -28,3 +28,4 @@ Clojure,clj,vector,list,vector,map,nil,true,false
 Dart,dart,list,list,record,map,null,true,false
 Hy,hy,list,list,tuple,dictionary,None,True,False
 Zig,zig,slice,slice,tuple,hash map,null,true,false
+Tcl,tcl,list,list,list,dictionary,empty string,true,false

evaluation/Makefile

@@ -1,10 +1,11 @@
 DOCKER_EXEC=podman
 
-build: Dockerfile Adb.dockerfile Hy.dockerfile Zig.dockerfile
+build: Dockerfile Adb.dockerfile Hy.dockerfile Zig.dockerfile Tcl.dockerfile
 	${DOCKER_EXEC} build -t multipl-e-evaluation .
 	${DOCKER_EXEC} build -t ghcr.io/nuprl/multipl-e:adb -f Adb.dockerfile .
 	${DOCKER_EXEC} build -t ghcr.io/nuprl/multipl-e:hy -f Hy.dockerfile .
 	${DOCKER_EXEC} build -t ghcr.io/nuprl/multipl-e:zig -f Zig.dockerfile .
+	${DOCKER_EXEC} build -t ghcr.io/nuprl/multipl-e:tcl -f Tcl.dockerfile .
 
 test: build
 	${DOCKER_EXEC} run --rm \

evaluation/Tcl.dockerfile

@@ -0,0 +1,9 @@
+FROM ubuntu:22.04
+ARG DEBIAN_FRONTEND=noninteractive
+ENV TZ=Etc/UTC
+
+RUN apt-get update -yqq && apt-get install -yqq python3-tqdm tcl
+
+COPY src /code
+WORKDIR /code
+ENTRYPOINT ["python3", "main.py"]

evaluation/src/containerized_eval.py

@@ -31,6 +31,7 @@
 import eval_dart
 import eval_hy
 import eval_zig
+import eval_tcl
 import tempfile
 
 
@@ -69,6 +70,7 @@
     "dart": (eval_dart.eval_script, ".dart"),
     "hy": (eval_hy.eval_script, ".hy"),
     "zig": (eval_zig.eval_script, ".zig"),
+    "tcl": (eval_tcl.eval_script, ".tcl"),
 }
 
 def eval_string_script(language, program):

evaluation/src/eval_tcl.py

@@ -0,0 +1,17 @@
+from pathlib import Path
+from safe_subprocess import run
+
+def eval_script(path: Path):
+    r = run(["tclsh", str(path)])
+    if r.timeout:
+        status = "Timeout"
+    elif r.exit_code == 0:
+        status = "OK"
+    else:
+        status = "Exception"
+    return {
+        "status": status,
+        "exit_code": r.exit_code,
+        "stdout": r.stdout,
+        "stderr": r.stderr,
+    }