scripts/apply_morph_to_urdf.py

#!/usr/bin/env python3
"""Apply Robot-Customization-Assets morph specs to an official URDF.

Stdlib only. Keep the scale rules in sync with
``unirobo.data_pipeline.config_catalog.write_joint_origin_urdf``.

Does **not** rewrite triangle-mesh scale (mplib/FCL rejects anisotropic mesh
scale). Only joint-origin ``xyz`` and prismatic gripper limits change.
"""
from __future__ import annotations

import argparse
import hashlib
import json
import os
import re
import sys
import xml.etree.ElementTree as ET
from pathlib import Path
from typing import Any

SKIP_LINK_TOKENS = (
    "world", "finger", "gripper", "knuckle", "camera", "wsg", "pad",
    "footprint", "inertial", "wheel", "castor", "caster",
)
SKIP_EXACT = frozenset({"base", "base_link", "ee_link", "tool0", "panda_link0"})
ARM_LINK_RE = re.compile(
    r"^(?:(?:fl_|fr_)link[1-6]|link[1-6]|panda_link[1-7]|"
    r"(?:shoulder|upper_arm|forearm|wrist_\d+)_link)$",
    re.I,
)
FINGER_WIDTH_MIN = 0.85
FINGER_WIDTH_MAX = 1.12


def _vec(text: str | None) -> list[float]:
    if not text:
        return [0.0, 0.0, 0.0]
    return [float(item) for item in text.replace(",", " ").split()]


def _fmt(values: list[float]) -> str:
    return " ".join(f"{float(item):.10g}" for item in values)


def _is_gripper_name(*parts: str | None) -> bool:
    blob = " ".join(part or "" for part in parts).lower()
    return any(token in blob for token in ("finger", "gripper", "knuckle", "wsg"))


def arm_links(urdf: Path) -> list[str]:
    names: list[str] = []
    fallback: list[str] = []
    for link in ET.parse(urdf).getroot().findall("link"):
        name = link.get("name") or ""
        lowered = name.lower()
        if ARM_LINK_RE.match(name):
            names.append(name)
            continue
        if lowered in SKIP_EXACT:
            continue
        if any(token in lowered for token in SKIP_LINK_TOKENS):
            continue
        fallback.append(name)
    if names:
        return names
    if not fallback:
        raise RuntimeError(f"no scalable arm links in {urdf}")
    return fallback


def link_scale_map(link_names: list[str], spec: dict[str, Any]) -> dict[str, float]:
    groups: dict[str, list[str]] = {}
    for name in link_names:
        if name.startswith("fl_"):
            groups.setdefault("fl", []).append(name)
        elif name.startswith("fr_"):
            groups.setdefault("fr", []).append(name)
        else:
            groups.setdefault("arm", []).append(name)
    link_scale: dict[str, float] = {}
    for members in groups.values():
        split = max(1, len(members) // 2)
        for index, name in enumerate(members):
            link_scale[name] = spec["proximal"] if index < split else spec["distal"]
    return link_scale


def _retarget_meshes(root: ET.Element, source_urdf: Path, output_urdf: Path) -> None:
    src_dir = source_urdf.parent.resolve()
    out_dir = output_urdf.parent.resolve()
    if src_dir == out_dir:
        return
    for mesh in root.findall(".//mesh"):
        filename = mesh.get("filename")
        if not filename or filename.startswith("package://"):
            continue
        path = Path(filename)
        if not path.is_absolute():
            path = (src_dir / path).resolve()
        try:
            mesh.set("filename", os.path.relpath(path, out_dir))
        except ValueError:
            mesh.set("filename", str(path))


def write_joint_origin_urdf(
    source_urdf: Path, output_urdf: Path, spec: dict[str, Any],
) -> dict[str, Any]:
    tree = ET.parse(source_urdf)
    root = tree.getroot()
    links = arm_links(source_urdf)
    scales = link_scale_map(links, spec)
    finger_length = float(spec["finger_length"])
    opening = min(FINGER_WIDTH_MAX, max(FINGER_WIDTH_MIN, float(spec["finger_width"])))
    joint_updates = []
    gripper_updates = []
    for joint in root.findall("joint"):
        parent = joint.find("parent")
        origin = joint.find("origin")
        parent_name = parent.get("link") if parent is not None else ""
        child = joint.find("child")
        child_name = child.get("link") if child is not None else ""
        name = joint.get("name") or ""
        gripper = _is_gripper_name(name, parent_name, child_name)
        scale = finger_length if gripper else float(scales.get(parent_name, 1.0))
        if origin is not None and origin.get("xyz") and scale != 1.0:
            before = _vec(origin.get("xyz"))
            after = [item * scale for item in before]
            origin.set("xyz", _fmt(after))
            joint_updates.append({
                "joint": name, "parent": parent_name, "scale": scale,
                "before_xyz": before, "after_xyz": after,
            })
        if gripper and joint.get("type") == "prismatic" and opening != 1.0:
            limit = joint.find("limit")
            if limit is not None:
                row: dict[str, Any] = {"joint": name, "opening_scale": opening}
                for bound in ("lower", "upper"):
                    if limit.get(bound) is not None:
                        before_v = float(limit.get(bound))
                        limit.set(bound, f"{before_v * opening:.10g}")
                        row[bound] = {"before": before_v, "after": before_v * opening}
                gripper_updates.append(row)
    for mesh in root.findall(".//mesh"):
        scale = mesh.get("scale")
        if scale and any(abs(float(item) - 1.0) > 1e-9 for item in _vec(scale)):
            raise RuntimeError(
                f"refusing non-unit mesh scale {scale} in {source_urdf}; "
                "mplib/FCL requires unit scale")
    _retarget_meshes(root, source_urdf, output_urdf)
    output_urdf.parent.mkdir(parents=True, exist_ok=True)
    tree.write(output_urdf, encoding="utf-8", xml_declaration=True)
    digest = hashlib.sha256(output_urdf.read_bytes()).hexdigest()
    return {
        "link_scale": scales,
        "finger_length_scale": finger_length,
        "opening_scale": opening,
        "n_joint_origin_updates": len(joint_updates),
        "n_gripper_limit_updates": len(gripper_updates),
        "mesh_scale_policy": "unchanged_unit",
        "output": str(output_urdf),
        "output_sha256": digest,
    }


def _load_specs(path: Path) -> dict[str, dict[str, Any]]:
    payload = json.loads(path.read_text(encoding="utf-8"))
    if not isinstance(payload, dict):
        raise ValueError(f"{path} must be an object of morph_id -> spec")
    return payload


def resolve_source_urdf(source: Path | None, embodiment_dir: Path | None) -> Path:
    if source is not None:
        if not source.is_file():
            raise FileNotFoundError(source)
        return source.resolve()
    if embodiment_dir is None:
        raise ValueError("pass --source or --embodiment-dir")
    config = embodiment_dir / "config.yml"
    if config.is_file():
        match = re.search(r"urdf_path:\s*(\S+)", config.read_text(encoding="utf-8"))
        if match:
            raw = match.group(1).strip().strip("\"'")
            raw = re.sub(r"^\$\{ASSETS_PATH\}/", "", raw)
            candidate = Path(raw)
            if candidate.is_file():
                return candidate.resolve()
            local = embodiment_dir / candidate.name
            if local.is_file():
                return local.resolve()
            nested = embodiment_dir / candidate
            if nested.is_file():
                return nested.resolve()
    found = sorted(embodiment_dir.glob("*.urdf"))
    if len(found) == 1:
        return found[0].resolve()
    raise FileNotFoundError(
        f"could not resolve a URDF under {embodiment_dir}; pass --source")


def main() -> int:
    here = Path(__file__).resolve()
    default_spec = here.parents[1] / "morph_specs" / "morphs.json"
    parser = argparse.ArgumentParser(description=__doc__)
    parser.add_argument("--source", type=Path, help="Official URDF path")
    parser.add_argument(
        "--embodiment-dir", type=Path,
        help="RoboTwin assets/embodiments/<name> (reads config.yml urdf_path)")
    parser.add_argument("--spec", type=Path, default=default_spec,
                        help="morphs.json from this dataset")
    parser.add_argument("--morph-id", help="One key from morphs.json")
    parser.add_argument("--all", action="store_true",
                        help="Write one URDF per morph id")
    parser.add_argument("--output", type=Path, help="Output URDF (single morph)")
    parser.add_argument("--output-dir", type=Path,
                        help="Directory for --all (default: next to the source URDF)")
    parser.add_argument("--list", action="store_true", help="Print morph ids")
    args = parser.parse_args()
    specs = _load_specs(args.spec)
    if args.list:
        for name, spec in specs.items():
            print(f"{name}\t{spec.get('description', '')}")
        return 0
    source = resolve_source_urdf(args.source, args.embodiment_dir)
    wanted = list(specs) if args.all else [args.morph_id]
    if not args.all and not args.morph_id:
        parser.error("pass --morph-id, --all, or --list")
    unknown = [name for name in wanted if name not in specs]
    if unknown:
        raise KeyError(f"unknown morph id: {unknown[:8]}")
    reports = []
    if args.all:
        out_dir = args.output_dir or source.parent
        for name in wanted:
            output = out_dir / f"{source.stem}__{name}{source.suffix}"
            reports.append(write_joint_origin_urdf(source, output, specs[name]))
    else:
        output = args.output
        if output is None:
            output = source.with_name(f"{source.stem}__{args.morph_id}{source.suffix}")
        reports.append(write_joint_origin_urdf(source, output, specs[args.morph_id]))
    print(json.dumps(reports if args.all else reports[0], indent=2))
    return 0


if __name__ == "__main__":
    raise SystemExit(main())