From 10f2bb93fa24f01b20f38bf03a8b76651d73ce45 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Peter=20Rowlands=20=28=EB=B3=80=EA=B8=B0=ED=98=B8=29?=
 <peter@pmrowla.com>
Date: Wed, 27 May 2020 18:07:38 +0900
Subject: [PATCH] ignore: CleanTree should always check ignores, not just for
 walk() (#3876)

* tests: test that CleanTree works outside of walk()

* ignore: CleanTree should always check ignores, not just for walk()

* CleanTree: handle broken symlinks

* handle case for linking from shared local cache dirs

* Fix windows relpath/different drive issue
---
 dvc/ignore.py           | 76 +++++++++++++++++++++++++++++++++++++----
 tests/func/test_tree.py | 21 ++++++++++++
 2 files changed, 91 insertions(+), 6 deletions(-)

diff --git a/dvc/ignore.py b/dvc/ignore.py
index eefa3bfef..384ba38bd 100644
--- a/dvc/ignore.py
+++ b/dvc/ignore.py
@@ -5,7 +5,9 @@ from funcy import cached_property
 from pathspec import PathSpec
 from pathspec.patterns import GitWildMatchPattern
 
+from dvc.path_info import PathInfo
 from dvc.scm.tree import BaseTree
+from dvc.utils import relpath
 
 logger = logging.getLogger(__name__)
 
@@ -125,19 +127,79 @@ class CleanTree(BaseTree):
         return self.tree.tree_root
 
     def open(self, path, mode="r", encoding="utf-8"):
-        return self.tree.open(path, mode, encoding)
+        if self.isfile(path):
+            return self.tree.open(path, mode, encoding)
+        raise FileNotFoundError
 
     def exists(self, path):
-        return self.tree.exists(path)
+        if self.tree.exists(path) and self._parents_exist(path):
+            if self.tree.isdir(path):
+                return self._valid_dirname(path)
+            return self._valid_filename(path)
+        return False
 
     def isdir(self, path):
-        return self.tree.isdir(path)
+        return (
+            self.tree.isdir(path)
+            and self._parents_exist(path)
+            and self._valid_dirname(path)
+        )
+
+    def _valid_dirname(self, path):
+        dirname, basename = os.path.split(os.path.normpath(path))
+        dirs, _ = self.dvcignore(os.path.abspath(dirname), [basename], [])
+        if dirs:
+            return True
+        return False
 
     def isfile(self, path):
-        return self.tree.isfile(path)
+        return (
+            self.tree.isfile(path)
+            and self._parents_exist(path)
+            and self._valid_filename(path)
+        )
+
+    def _valid_filename(self, path):
+        dirname, basename = os.path.split(os.path.normpath(path))
+        _, files = self.dvcignore(os.path.abspath(dirname), [], [basename])
+        if files:
+            return True
+        return False
 
     def isexec(self, path):
-        return self.tree.isexec(path)
+        return self.exists(path) and self.tree.isexec(path)
+
+    def _parents_exist(self, path):
+        from dvc.repo import Repo
+
+        path = PathInfo(path)
+
+        # if parent is tree_root or inside a .dvc dir we can skip this check
+        if path.parent == self.tree_root or Repo.DVC_DIR in path.parts:
+            return True
+
+        # if path is outside of tree, assume this is a local remote/local cache
+        # link/move operation where we do not need to filter ignores
+        path = relpath(path, self.tree_root)
+        if path.startswith("..") or (
+            os.name == "nt"
+            and not os.path.commonprefix(
+                [os.path.abspath(path), self.tree_root]
+            )
+        ):
+            return True
+
+        # check if parent directories are in our ignores, starting from
+        # tree_root
+        for parent_dir in reversed(PathInfo(path).parents):
+            dirname, basename = os.path.split(parent_dir)
+            if basename == ".":
+                # parent_dir == tree_root
+                continue
+            dirs, _ = self.dvcignore(os.path.abspath(dirname), [basename], [])
+            if not dirs:
+                return False
+        return True
 
     def walk(self, top, topdown=True):
         for root, dirs, files in self.tree.walk(top, topdown):
@@ -148,4 +210,6 @@ class CleanTree(BaseTree):
             yield root, dirs, files
 
     def stat(self, path):
-        return self.tree.stat(path)
+        if self.exists(path):
+            return self.tree.stat(path)
+        raise FileNotFoundError
diff --git a/tests/func/test_tree.py b/tests/func/test_tree.py
index 64bb733aa..1df0000d5 100644
--- a/tests/func/test_tree.py
+++ b/tests/func/test_tree.py
@@ -3,6 +3,7 @@ from os.path import join
 
 from dvc.ignore import CleanTree
 from dvc.path_info import PathInfo
+from dvc.repo import Repo
 from dvc.repo.tree import RepoTree
 from dvc.scm import SCM
 from dvc.scm.git import GitTree
@@ -220,3 +221,23 @@ def test_repotree_cache_save(tmp_dir, dvc, scm, erepo_dir, setup_remote):
             cache.save(PathInfo(erepo_dir / "dir"), None, tree=tree)
     for checksum in expected:
         assert os.path.exists(cache.checksum_to_path_info(checksum))
+
+
+def test_cleantree_subrepo(tmp_dir, dvc, scm, monkeypatch):
+    tmp_dir.gen({"subdir": {}})
+    subrepo_dir = tmp_dir / "subdir"
+    with subrepo_dir.chdir():
+        subrepo = Repo.init(subdir=True)
+        subrepo_dir.gen({"foo": "foo", "dir": {"bar": "bar"}})
+
+    assert isinstance(dvc.tree, CleanTree)
+    assert not dvc.tree.exists(subrepo_dir / "foo")
+    assert not dvc.tree.isfile(subrepo_dir / "foo")
+    assert not dvc.tree.exists(subrepo_dir / "dir")
+    assert not dvc.tree.isdir(subrepo_dir / "dir")
+
+    assert isinstance(subrepo.tree, CleanTree)
+    assert subrepo.tree.exists(subrepo_dir / "foo")
+    assert subrepo.tree.isfile(subrepo_dir / "foo")
+    assert subrepo.tree.exists(subrepo_dir / "dir")
+    assert subrepo.tree.isdir(subrepo_dir / "dir")