diff --git a/swh/model/from_disk.py b/swh/model/from_disk.py index da55912..8551fdd 100644 --- a/swh/model/from_disk.py +++ b/swh/model/from_disk.py @@ -1,448 +1,448 @@ # Copyright (C) 2017-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import datetime import enum import os import stat import attr from attrs_strict import type_validator from typing import List, Optional, Iterable, Any from typing_extensions import Final from .hashutil import MultiHash from .merkle import MerkleLeaf, MerkleNode from .identifiers import ( directory_entry_sort_key, directory_identifier, identifier_to_bytes as id_to_bytes, identifier_to_str as id_to_str, ) from . import model @attr.s class DiskBackedContent(model.BaseContent): """Content-like class, which allows lazy-loading data from the disk.""" object_type: Final = "content_file" sha1 = attr.ib(type=bytes, validator=type_validator()) sha1_git = attr.ib(type=model.Sha1Git, validator=type_validator()) sha256 = attr.ib(type=bytes, validator=type_validator()) blake2s256 = attr.ib(type=bytes, validator=type_validator()) length = attr.ib(type=int, validator=type_validator()) status = attr.ib( type=str, validator=attr.validators.in_(["visible", "hidden"]), default="visible", ) ctime = attr.ib( type=Optional[datetime.datetime], validator=type_validator(), default=None, eq=False, ) path = attr.ib(type=Optional[bytes], default=None) @classmethod def from_dict(cls, d): return cls(**d) def __attrs_post_init__(self): if self.path is None: raise TypeError("path must not be None.") def with_data(self) -> model.Content: args = self.to_dict() del args["path"] assert self.path is not None with open(self.path, "rb") as fd: return model.Content.from_dict({**args, "data": fd.read()}) class DentryPerms(enum.IntEnum): """Admissible permissions for directory entries.""" content = 0o100644 """Content""" executable_content = 0o100755 """Executable content (e.g. executable script)""" symlink = 0o120000 """Symbolic link""" directory = 0o040000 """Directory""" revision = 0o160000 """Revision (e.g. submodule)""" def mode_to_perms(mode): """Convert a file mode to a permission compatible with Software Heritage directory entries Args: mode (int): a file mode as returned by :func:`os.stat` in :attr:`os.stat_result.st_mode` Returns: DentryPerms: one of the following values: :const:`DentryPerms.content`: plain file :const:`DentryPerms.executable_content`: executable file :const:`DentryPerms.symlink`: symbolic link :const:`DentryPerms.directory`: directory """ if stat.S_ISLNK(mode): return DentryPerms.symlink if stat.S_ISDIR(mode): return DentryPerms.directory else: # file is executable in any way if mode & (0o111): return DentryPerms.executable_content else: return DentryPerms.content class Content(MerkleLeaf): """Representation of a Software Heritage content as a node in a Merkle tree. The current Merkle hash for the Content nodes is the `sha1_git`, which makes it consistent with what :class:`Directory` uses for its own hash computation. """ __slots__ = [] # type: List[str] - type = "content" + object_type: Final = "content" @classmethod def from_bytes(cls, *, mode, data): """Convert data (raw :class:`bytes`) to a Software Heritage content entry Args: mode (int): a file mode (passed to :func:`mode_to_perms`) data (bytes): raw contents of the file """ ret = MultiHash.from_data(data).digest() ret["length"] = len(data) ret["perms"] = mode_to_perms(mode) ret["data"] = data ret["status"] = "visible" return cls(ret) @classmethod def from_symlink(cls, *, path, mode): """Convert a symbolic link to a Software Heritage content entry""" return cls.from_bytes(mode=mode, data=os.readlink(path)) @classmethod def from_file(cls, *, path, max_content_length=None): """Compute the Software Heritage content entry corresponding to an on-disk file. The returned dictionary contains keys useful for both: - loading the content in the archive (hashes, `length`) - using the content as a directory entry in a directory Args: save_path (bool): add the file path to the entry max_content_length (Optional[int]): if given, all contents larger than this will be skipped. """ file_stat = os.lstat(path) mode = file_stat.st_mode length = file_stat.st_size too_large = max_content_length is not None and length > max_content_length if stat.S_ISLNK(mode): # Symbolic link: return a file whose contents are the link target if too_large: # Unlike large contents, we can't stream symlinks to # MultiHash, and we don't want to fit them in memory if # they exceed max_content_length either. # Thankfully, this should not happen for reasonable values of # max_content_length because of OS/filesystem limitations, # so let's just raise an error. raise Exception(f"Symlink too large ({length} bytes)") return cls.from_symlink(path=path, mode=mode) elif not stat.S_ISREG(mode): # not a regular file: return the empty file instead return cls.from_bytes(mode=mode, data=b"") if too_large: skip_reason = "Content too large" else: skip_reason = None hashes = MultiHash.from_path(path).digest() if skip_reason: ret = { **hashes, "status": "absent", "reason": skip_reason, } else: ret = { **hashes, "status": "visible", } ret["path"] = path ret["perms"] = mode_to_perms(mode) ret["length"] = length obj = cls(ret) return obj def __repr__(self): return "Content(id=%s)" % id_to_str(self.hash) def compute_hash(self): return self.data["sha1_git"] def to_model(self) -> model.BaseContent: """Builds a `model.BaseContent` object based on this leaf.""" data = self.get_data().copy() data.pop("perms", None) if data["status"] == "absent": data.pop("path", None) return model.SkippedContent.from_dict(data) elif "data" in data: return model.Content.from_dict(data) else: return DiskBackedContent.from_dict(data) def accept_all_directories(dirpath: str, dirname: str, entries: Iterable[Any]) -> bool: """Default filter for :func:`Directory.from_disk` accepting all directories Args: dirname (bytes): directory name entries (list): directory entries """ return True def ignore_empty_directories( dirpath: str, dirname: str, entries: Iterable[Any] ) -> bool: """Filter for :func:`directory_to_objects` ignoring empty directories Args: dirname (bytes): directory name entries (list): directory entries Returns: True if the directory is not empty, false if the directory is empty """ return bool(entries) def ignore_named_directories(names, *, case_sensitive=True): """Filter for :func:`directory_to_objects` to ignore directories named one of names. Args: names (list of bytes): names to ignore case_sensitive (bool): whether to do the filtering in a case sensitive way Returns: a directory filter for :func:`directory_to_objects` """ if not case_sensitive: names = [name.lower() for name in names] def named_filter( dirpath: str, dirname: str, entries: Iterable[Any], names: Iterable[Any] = names, case_sensitive: bool = case_sensitive, ): if case_sensitive: return dirname not in names else: return dirname.lower() not in names return named_filter class Directory(MerkleNode): """Representation of a Software Heritage directory as a node in a Merkle Tree. This class can be used to generate, from an on-disk directory, all the objects that need to be sent to the Software Heritage archive. The :func:`from_disk` constructor allows you to generate the data structure from a directory on disk. The resulting :class:`Directory` can then be manipulated as a dictionary, using the path as key. The :func:`collect` method is used to retrieve all the objects that need to be added to the Software Heritage archive since the last collection, by class (contents and directories). When using the dict-like methods to update the contents of the directory, the affected levels of hierarchy are reset and can be collected again using the same method. This enables the efficient collection of updated nodes, for instance when the client is applying diffs. """ __slots__ = ["__entries"] - type = "directory" + object_type: Final = "directory" @classmethod def from_disk( cls, *, path, dir_filter=accept_all_directories, max_content_length=None ): """Compute the Software Heritage objects for a given directory tree Args: path (bytes): the directory to traverse data (bool): whether to add the data to the content objects save_path (bool): whether to add the path to the content objects dir_filter (function): a filter to ignore some directories by name or contents. Takes two arguments: dirname and entries, and returns True if the directory should be added, False if the directory should be ignored. max_content_length (Optional[int]): if given, all contents larger than this will be skipped. """ top_path = path dirs = {} for root, dentries, fentries in os.walk(top_path, topdown=False): entries = {} # Join fentries and dentries in the same processing, as symbolic # links to directories appear in dentries... for name in fentries + dentries: path = os.path.join(root, name) if not os.path.isdir(path) or os.path.islink(path): content = Content.from_file( path=path, max_content_length=max_content_length ) entries[name] = content else: if dir_filter(path, name, dirs[path].entries): entries[name] = dirs[path] dirs[root] = cls({"name": os.path.basename(root)}) dirs[root].update(entries) return dirs[top_path] def __init__(self, data=None): super().__init__(data=data) self.__entries = None def invalidate_hash(self): self.__entries = None super().invalidate_hash() @staticmethod def child_to_directory_entry(name, child): - if isinstance(child, Directory): + if child.object_type == "directory": return { "type": "dir", "perms": DentryPerms.directory, "target": child.hash, "name": name, } - elif isinstance(child, Content): + elif child.object_type == "content": return { "type": "file", "perms": child.data["perms"], "target": child.hash, "name": name, } else: - raise ValueError("unknown child") + raise ValueError(f"unknown child {child}") def get_data(self, **kwargs): return { "id": self.hash, "entries": self.entries, } @property def entries(self): """Child nodes, sorted by name in the same way `directory_identifier` does.""" if self.__entries is None: self.__entries = sorted( ( self.child_to_directory_entry(name, child) for name, child in self.items() ), key=directory_entry_sort_key, ) return self.__entries def compute_hash(self): return id_to_bytes(directory_identifier({"entries": self.entries})) def to_model(self) -> model.Directory: """Builds a `model.Directory` object based on this node; ignoring its children.""" return model.Directory.from_dict(self.get_data()) def __getitem__(self, key): if not isinstance(key, bytes): raise ValueError("Can only get a bytes from Directory") # Convenience shortcut if key == b"": return self if b"/" not in key: return super().__getitem__(key) else: key1, key2 = key.split(b"/", 1) return self.__getitem__(key1)[key2] def __setitem__(self, key, value): if not isinstance(key, bytes): raise ValueError("Can only set a bytes Directory entry") if not isinstance(value, (Content, Directory)): raise ValueError( "Can only set a Directory entry to a Content or " "Directory" ) if key == b"": raise ValueError("Directory entry must have a name") if b"\x00" in key: raise ValueError("Directory entry name must not contain nul bytes") if b"/" not in key: return super().__setitem__(key, value) else: key1, key2 = key.rsplit(b"/", 1) self[key1].__setitem__(key2, value) def __delitem__(self, key): if not isinstance(key, bytes): raise ValueError("Can only delete a bytes Directory entry") if b"/" not in key: super().__delitem__(key) else: key1, key2 = key.rsplit(b"/", 1) del self[key1][key2] def __repr__(self): return "Directory(id=%s, entries=[%s])" % ( id_to_str(self.hash), ", ".join(str(entry) for entry in self), ) diff --git a/swh/model/merkle.py b/swh/model/merkle.py index 36407d8..bddaa3c 100644 --- a/swh/model/merkle.py +++ b/swh/model/merkle.py @@ -1,313 +1,312 @@ -# Copyright (C) 2017 The Software Heritage developers +# Copyright (C) 2017-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information """Merkle tree data structure""" import abc import collections -from typing import Iterator, List, Optional, Set +from typing import Iterator, List, Set def deep_update(left, right): """Recursively update the left mapping with deeply nested values from the right mapping. This function is useful to merge the results of several calls to :func:`MerkleNode.collect`. Arguments: left: a mapping (modified by the update operation) right: a mapping Returns: the left mapping, updated with nested values from the right mapping Example: >>> a = { ... 'key1': { ... 'key2': { ... 'key3': 'value1/2/3', ... }, ... }, ... } >>> deep_update(a, { ... 'key1': { ... 'key2': { ... 'key4': 'value1/2/4', ... }, ... }, ... }) == { ... 'key1': { ... 'key2': { ... 'key3': 'value1/2/3', ... 'key4': 'value1/2/4', ... }, ... }, ... } True >>> deep_update(a, { ... 'key1': { ... 'key2': { ... 'key3': 'newvalue1/2/3', ... }, ... }, ... }) == { ... 'key1': { ... 'key2': { ... 'key3': 'newvalue1/2/3', ... 'key4': 'value1/2/4', ... }, ... }, ... } True """ for key, rvalue in right.items(): if isinstance(rvalue, collections.Mapping): new_lvalue = deep_update(left.get(key, {}), rvalue) left[key] = new_lvalue else: left[key] = rvalue return left class MerkleNode(dict, metaclass=abc.ABCMeta): """Representation of a node in a Merkle Tree. A (generalized) `Merkle Tree`_ is a tree in which every node is labeled with a hash of its own data and the hash of its children. .. _Merkle Tree: https://en.wikipedia.org/wiki/Merkle_tree In pseudocode:: node.hash = hash(node.data + sum(child.hash for child in node.children)) This class efficiently implements the Merkle Tree data structure on top of a Python :class:`dict`, minimizing hash computations and new data collections when updating nodes. Node data is stored in the :attr:`data` attribute, while (named) children are stored as items of the underlying dictionary. Addition, update and removal of objects are instrumented to automatically invalidate the hashes of the current node as well as its registered parents; It also resets the collection status of the objects so the updated objects can be collected. The collection of updated data from the tree is implemented through the :func:`collect` function and associated helpers. Attributes: data (dict): data associated to the current node parents (list): known parents of the current node collected (bool): whether the current node has been collected """ __slots__ = ["parents", "data", "__hash", "collected"] - type = None # type: Optional[str] # TODO: make this an enum """Type of the current node (used as a classifier for :func:`collect`)""" def __init__(self, data=None): super().__init__() self.parents = [] self.data = data self.__hash = None self.collected = False def __eq__(self, other): return ( isinstance(other, MerkleNode) and super().__eq__(other) and self.data == other.data ) def __ne__(self, other): return not self.__eq__(other) def invalidate_hash(self): """Invalidate the cached hash of the current node.""" if not self.__hash: return self.__hash = None self.collected = False for parent in self.parents: parent.invalidate_hash() def update_hash(self, *, force=False): """Recursively compute the hash of the current node. Args: force (bool): invalidate the cache and force the computation for this node and all children. """ if self.__hash and not force: return self.__hash if force: self.invalidate_hash() for child in self.values(): child.update_hash(force=force) self.__hash = self.compute_hash() return self.__hash @property def hash(self): """The hash of the current node, as calculated by :func:`compute_hash`. """ return self.update_hash() @abc.abstractmethod def compute_hash(self): """Compute the hash of the current node. The hash should depend on the data of the node, as well as on hashes of the children nodes. """ raise NotImplementedError("Must implement compute_hash method") def __setitem__(self, name, new_child): """Add a child, invalidating the current hash""" self.invalidate_hash() super().__setitem__(name, new_child) new_child.parents.append(self) def __delitem__(self, name): """Remove a child, invalidating the current hash""" if name in self: self.invalidate_hash() self[name].parents.remove(self) super().__delitem__(name) else: raise KeyError(name) def update(self, new_children): """Add several named children from a dictionary""" if not new_children: return self.invalidate_hash() for name, new_child in new_children.items(): new_child.parents.append(self) if name in self: self[name].parents.remove(self) super().update(new_children) def get_data(self, **kwargs): """Retrieve and format the collected data for the current node, for use by :func:`collect`. Can be overridden, for instance when you want the collected data to contain information about the child nodes. Arguments: kwargs: allow subclasses to alter behaviour depending on how :func:`collect` is called. Returns: data formatted for :func:`collect` """ return self.data def collect_node(self, **kwargs): """Collect the data for the current node, for use by :func:`collect`. Arguments: kwargs: passed as-is to :func:`get_data`. Returns: A :class:`dict` compatible with :func:`collect`. """ if not self.collected: self.collected = True - return {self.type: {self.hash: self.get_data(**kwargs)}} + return {self.object_type: {self.hash: self.get_data(**kwargs)}} else: return {} def collect(self, **kwargs): """Collect the data for all nodes in the subtree rooted at `self`. The data is deduplicated by type and by hash. Arguments: kwargs: passed as-is to :func:`get_data`. Returns: A :class:`dict` with the following structure:: { 'typeA': { node1.hash: node1.get_data(), node2.hash: node2.get_data(), }, 'typeB': { node3.hash: node3.get_data(), ... }, ... } """ ret = self.collect_node(**kwargs) for child in self.values(): deep_update(ret, child.collect(**kwargs)) return ret def reset_collect(self): """Recursively unmark collected nodes in the subtree rooted at `self`. This lets the caller use :func:`collect` again. """ self.collected = False for child in self.values(): child.reset_collect() def iter_tree(self) -> Iterator["MerkleNode"]: """Yields all children nodes, recursively. Common nodes are deduplicated. """ yield from self._iter_tree(set()) def _iter_tree(self, seen: Set[bytes]) -> Iterator["MerkleNode"]: if self.hash not in seen: seen.add(self.hash) yield self for child in self.values(): yield from child._iter_tree(seen=seen) class MerkleLeaf(MerkleNode): """A leaf to a Merkle tree. A Merkle leaf is simply a Merkle node with children disabled. """ __slots__ = [] # type: List[str] def __setitem__(self, name, child): raise ValueError("%s is a leaf" % self.__class__.__name__) def __getitem__(self, name): raise ValueError("%s is a leaf" % self.__class__.__name__) def __delitem__(self, name): raise ValueError("%s is a leaf" % self.__class__.__name__) def update(self, new_children): """Children update operation. Disabled for leaves.""" raise ValueError("%s is a leaf" % self.__class__.__name__) diff --git a/swh/model/tests/test_merkle.py b/swh/model/tests/test_merkle.py index 7e3538b..65992f4 100644 --- a/swh/model/tests/test_merkle.py +++ b/swh/model/tests/test_merkle.py @@ -1,243 +1,247 @@ -# Copyright (C) 2017 The Software Heritage developers +# Copyright (C) 2017-2020 The Software Heritage developers # See the AUTHORS file at the top-level directory of this distribution # License: GNU General Public License version 3, or any later version # See top-level LICENSE file for more information import unittest from swh.model import merkle class MerkleTestNode(merkle.MerkleNode): - type = "tested_merkle_node_type" + object_type = "tested_merkle_node_type" def __init__(self, data): super().__init__(data) self.compute_hash_called = 0 def compute_hash(self): self.compute_hash_called += 1 child_data = [child + b"=" + self[child].hash for child in sorted(self)] return b"hash(" + b", ".join([self.data["value"]] + child_data) + b")" class MerkleTestLeaf(merkle.MerkleLeaf): - type = "tested_merkle_leaf_type" + object_type = "tested_merkle_leaf_type" def __init__(self, data): super().__init__(data) self.compute_hash_called = 0 def compute_hash(self): self.compute_hash_called += 1 return b"hash(" + self.data["value"] + b")" class TestMerkleLeaf(unittest.TestCase): def setUp(self): self.data = {"value": b"value"} self.instance = MerkleTestLeaf(self.data) def test_equality(self): leaf1 = MerkleTestLeaf(self.data) leaf2 = MerkleTestLeaf(self.data) leaf3 = MerkleTestLeaf({}) self.assertEqual(leaf1, leaf2) self.assertNotEqual(leaf1, leaf3) def test_hash(self): self.assertEqual(self.instance.compute_hash_called, 0) instance_hash = self.instance.hash self.assertEqual(self.instance.compute_hash_called, 1) instance_hash2 = self.instance.hash self.assertEqual(self.instance.compute_hash_called, 1) self.assertEqual(instance_hash, instance_hash2) def test_data(self): self.assertEqual(self.instance.get_data(), self.data) def test_collect(self): collected = self.instance.collect() self.assertEqual( collected, - {self.instance.type: {self.instance.hash: self.instance.get_data(),},}, + { + self.instance.object_type: { + self.instance.hash: self.instance.get_data(), + }, + }, ) collected2 = self.instance.collect() self.assertEqual(collected2, {}) self.instance.reset_collect() collected3 = self.instance.collect() self.assertEqual(collected, collected3) def test_leaf(self): with self.assertRaisesRegex(ValueError, "is a leaf"): self.instance[b"key1"] = "Test" with self.assertRaisesRegex(ValueError, "is a leaf"): del self.instance[b"key1"] with self.assertRaisesRegex(ValueError, "is a leaf"): self.instance[b"key1"] with self.assertRaisesRegex(ValueError, "is a leaf"): self.instance.update(self.data) class TestMerkleNode(unittest.TestCase): maxDiff = None def setUp(self): self.root = MerkleTestNode({"value": b"root"}) self.nodes = {b"root": self.root} for i in (b"a", b"b", b"c"): value = b"root/" + i node = MerkleTestNode({"value": value,}) self.root[i] = node self.nodes[value] = node for j in (b"a", b"b", b"c"): value2 = value + b"/" + j node2 = MerkleTestNode({"value": value2,}) node[j] = node2 self.nodes[value2] = node2 for k in (b"a", b"b", b"c"): value3 = value2 + b"/" + j node3 = MerkleTestNode({"value": value3,}) node2[j] = node3 self.nodes[value3] = node3 def test_equality(self): node1 = merkle.MerkleNode({"foo": b"bar"}) node2 = merkle.MerkleNode({"foo": b"bar"}) node3 = merkle.MerkleNode({}) self.assertEqual(node1, node2) self.assertNotEqual(node1, node3, node1 == node3) node1["foo"] = node3 self.assertNotEqual(node1, node2) node2["foo"] = node3 self.assertEqual(node1, node2) def test_hash(self): for node in self.nodes.values(): self.assertEqual(node.compute_hash_called, 0) # Root hash will compute hash for all the nodes hash = self.root.hash for node in self.nodes.values(): self.assertEqual(node.compute_hash_called, 1) self.assertIn(node.data["value"], hash) # Should use the cached value hash2 = self.root.hash self.assertEqual(hash, hash2) for node in self.nodes.values(): self.assertEqual(node.compute_hash_called, 1) # Should still use the cached value hash3 = self.root.update_hash(force=False) self.assertEqual(hash, hash3) for node in self.nodes.values(): self.assertEqual(node.compute_hash_called, 1) # Force update of the cached value for a deeply nested node self.root[b"a"][b"b"].update_hash(force=True) for key, node in self.nodes.items(): # update_hash rehashes all children if key.startswith(b"root/a/b"): self.assertEqual(node.compute_hash_called, 2) else: self.assertEqual(node.compute_hash_called, 1) hash4 = self.root.hash self.assertEqual(hash, hash4) for key, node in self.nodes.items(): # update_hash also invalidates all parents if key in (b"root", b"root/a") or key.startswith(b"root/a/b"): self.assertEqual(node.compute_hash_called, 2) else: self.assertEqual(node.compute_hash_called, 1) def test_collect(self): collected = self.root.collect() - self.assertEqual(len(collected[self.root.type]), len(self.nodes)) + self.assertEqual(len(collected[self.root.object_type]), len(self.nodes)) for node in self.nodes.values(): self.assertTrue(node.collected) collected2 = self.root.collect() self.assertEqual(collected2, {}) def test_iter_tree(self): nodes = list(self.root.iter_tree()) self.assertCountEqual(nodes, self.nodes.values()) def test_get(self): for key in (b"a", b"b", b"c"): self.assertEqual(self.root[key], self.nodes[b"root/" + key]) with self.assertRaisesRegex(KeyError, "b'nonexistent'"): self.root[b"nonexistent"] def test_del(self): hash_root = self.root.hash hash_a = self.nodes[b"root/a"].hash del self.root[b"a"][b"c"] hash_root2 = self.root.hash hash_a2 = self.nodes[b"root/a"].hash self.assertNotEqual(hash_root, hash_root2) self.assertNotEqual(hash_a, hash_a2) self.assertEqual(self.nodes[b"root/a/c"].parents, []) with self.assertRaisesRegex(KeyError, "b'nonexistent'"): del self.root[b"nonexistent"] def test_update(self): hash_root = self.root.hash hash_b = self.root[b"b"].hash new_children = { b"c": MerkleTestNode({"value": b"root/b/new_c"}), b"d": MerkleTestNode({"value": b"root/b/d"}), } # collect all nodes self.root.collect() self.root[b"b"].update(new_children) # Ensure everyone got reparented self.assertEqual(new_children[b"c"].parents, [self.root[b"b"]]) self.assertEqual(new_children[b"d"].parents, [self.root[b"b"]]) self.assertEqual(self.nodes[b"root/b/c"].parents, []) hash_root2 = self.root.hash self.assertNotEqual(hash_root, hash_root2) self.assertIn(b"root/b/new_c", hash_root2) self.assertIn(b"root/b/d", hash_root2) hash_b2 = self.root[b"b"].hash self.assertNotEqual(hash_b, hash_b2) for key, node in self.nodes.items(): if key in (b"root", b"root/b"): self.assertEqual(node.compute_hash_called, 2) else: self.assertEqual(node.compute_hash_called, 1) # Ensure we collected root, root/b, and both new children collected_after_update = self.root.collect() self.assertCountEqual( - collected_after_update[MerkleTestNode.type], + collected_after_update[MerkleTestNode.object_type], [ self.nodes[b"root"].hash, self.nodes[b"root/b"].hash, new_children[b"c"].hash, new_children[b"d"].hash, ], ) # test that noop updates doesn't invalidate anything self.root[b"a"][b"b"].update({}) self.assertEqual(self.root.collect(), {})