stage collection: do not collect .gitignored dvcfiles (#5265)

* Do not collect .gitignored dvcfiles

* fix exc

* consistent error message

* fix imports

* add tests

dvc/dvcfile.py

@@ -37,6 +37,11 @@ PIPELINE_FILE = "dvc.yaml"
 PIPELINE_LOCK = "dvc.lock"
 
 
+class FileIsGitIgnored(DvcException):
+    def __init__(self, path):
+        super().__init__(f"'{path}' is git-ignored.")
+
+
 class LockfileCorruptedError(DvcException):
     pass
 
@@ -103,18 +108,29 @@ class FileMixin:
     def exists(self):
         return self.repo.tree.exists(self.path)
 
+    def _is_git_ignored(self):
+        from dvc.tree.local import LocalTree
+
+        return isinstance(
+            self.repo.tree, LocalTree
+        ) and self.repo.scm.is_ignored(self.path)
+
     def _load(self):
         # it raises the proper exceptions by priority:
         # 1. when the file doesn't exists
         # 2. filename is not a DVC-file
         # 3. path doesn't represent a regular file
+        # 4. when the file is git ignored
         if not self.exists():
-            raise StageFileDoesNotExistError(self.path)
+            is_ignored = self.repo.tree.exists(self.path, use_dvcignore=False)
+            raise StageFileDoesNotExistError(self.path, dvc_ignored=is_ignored)
+
         if self.verify:
             check_dvc_filename(self.path)
         if not self.repo.tree.isfile(self.path):
             raise StageFileIsNotDvcFileError(self.path)
-
+        if self._is_git_ignored():
+            raise FileIsGitIgnored(self.path)
         with self.repo.tree.open(self.path) as fd:
             stage_text = fd.read()
         d = parse_yaml(stage_text, self.path)

dvc/repo/stage.py

@@ -3,6 +3,7 @@ import logging
 import os
 import typing
 from contextlib import suppress
+from functools import partial
 from typing import (
     Callable,
     Iterable,
@@ -371,13 +372,33 @@ class StageLoad:
                 the collection.
         """
         from dvc.dvcfile import is_valid_filename
+        from dvc.tree.local import LocalTree
+
+        scm = self.repo.scm
+        sep = os.sep
+        outs: Set[str] = set()
+
+        is_local_tree = isinstance(self.tree, LocalTree)
+
+        def is_ignored(path):
+            # apply only for the local tree
+            return is_local_tree and scm.is_ignored(path)
+
+        def is_dvcfile_and_not_ignored(root, file):
+            return is_valid_filename(file) and not is_ignored(
+                f"{root}{sep}{file}"
+            )
+
+        def is_out_or_ignored(root, directory):
+            dir_path = f"{root}{sep}{directory}"
+            return dir_path in outs or is_ignored(dir_path)
 
         stages = []
-        outs: Set[str] = set()
         for root, dirs, files in self.tree.walk(self.repo.root_dir):
-            for file_name in filter(is_valid_filename, files):
-                file_path = os.path.join(root, file_name)
+            dvcfile_filter = partial(is_dvcfile_and_not_ignored, root)
 
+            for file in filter(dvcfile_filter, files):
+                file_path = os.path.join(root, file)
                 try:
                     new_stages = self.load_file(file_path)
                 except DvcException as exc:
@@ -393,5 +414,5 @@ class StageLoad:
                     for out in stage.outs
                     if out.scheme == "local"
                 )
-            dirs[:] = [d for d in dirs if os.path.join(root, d) not in outs]
+            dirs[:] = [d for d in dirs if not is_out_or_ignored(root, d)]
         return stages

dvc/stage/exceptions.py

@@ -14,9 +14,13 @@ class StageFileFormatError(DvcException):
 
 
 class StageFileDoesNotExistError(DvcException):
-    def __init__(self, fname):
+    DVC_IGNORED = "is dvc-ignored"
+    DOES_NOT_EXIST = "does not exist"
+
+    def __init__(self, fname, dvc_ignored=False):
         self.file = fname
-        super().__init__(f"'{self.file}' does not exist.")
+        message = self.DVC_IGNORED if dvc_ignored else self.DOES_NOT_EXIST
+        super().__init__(f"'{self.file}' {message}")
 
 
 class StageFileAlreadyExistsError(DvcException):

tests/func/test_stage_load.py

@@ -4,7 +4,7 @@ from operator import itemgetter
 import pytest
 from funcy import raiser
 
-from dvc.dvcfile import PIPELINE_FILE
+from dvc.dvcfile import PIPELINE_FILE, FileIsGitIgnored
 from dvc.exceptions import NoOutputOrStageError
 from dvc.path_info import PathInfo
 from dvc.repo import Repo
@@ -451,3 +451,45 @@ def test_collect_repo_callback(tmp_dir, dvc, mocker):
     file_path, exc = mock.call_args[0]
     assert file_path == PIPELINE_FILE
     assert isinstance(exc, StageFileFormatError)
+
+
+def test_gitignored_collect_repo(tmp_dir, dvc, scm):
+    (stage,) = tmp_dir.dvc_gen({"data": {"foo": "foo", "bar": "bar"}})
+
+    assert dvc.stage.collect_repo() == [stage]
+
+    scm.ignore(stage.path)
+    scm._reset()
+
+    assert not dvc.stage.collect_repo()
+
+
+def test_gitignored_file_try_collect_granular_for_data_files(
+    tmp_dir, dvc, scm
+):
+    (stage,) = tmp_dir.dvc_gen({"data": {"foo": "foo", "bar": "bar"}})
+    path = PathInfo("data") / "foo"
+
+    assert dvc.stage.collect_granular(str(path)) == [
+        (stage, PathInfo(tmp_dir / path))
+    ]
+
+    scm.ignore(stage.path)
+    dvc._reset()
+
+    with pytest.raises(NoOutputOrStageError):
+        dvc.stage.collect_granular(str(path))
+
+
+def test_gitignored_file_try_collect_granular_for_dvc_yaml_files(
+    tmp_dir, dvc, scm, stages
+):
+    assert dvc.stage.collect_granular("bar") == [
+        (stages["copy-foo-bar"], PathInfo(tmp_dir / "bar"))
+    ]
+
+    scm.ignore(tmp_dir / "dvc.yaml")
+    scm._reset()
+
+    with pytest.raises(FileIsGitIgnored):
+        dvc.stage.collect_granular("bar")

tests/unit/test_dvcfile.py

@@ -4,6 +4,7 @@ from dvc.dvcfile import (
     PIPELINE_FILE,
     PIPELINE_LOCK,
     Dvcfile,
+    FileIsGitIgnored,
     PipelineFile,
     SingleStageFile,
 )
@@ -41,12 +42,23 @@ def test_pipelines_single_stage_file(path):
 
 
 @pytest.mark.parametrize("file", ["stage.dvc", "dvc.yaml"])
-def test_stage_load_on_not_existing_file(tmp_dir, dvc, file):
+@pytest.mark.parametrize("is_dvcignored", [True, False])
+def test_stage_load_on_not_existing_file(tmp_dir, dvc, file, is_dvcignored):
     dvcfile = Dvcfile(dvc, file)
+    if is_dvcignored:
+        (tmp_dir / ".dvcignore").write_text(file)
+
     assert not dvcfile.exists()
-    with pytest.raises(StageFileDoesNotExistError):
+    with pytest.raises(StageFileDoesNotExistError) as exc_info:
         assert dvcfile.stages.values()
+
+    assert str(exc_info.value) == f"'{file}' does not exist"
+
+
+@pytest.mark.parametrize("file", ["stage.dvc", "dvc.yaml"])
+def test_stage_load_on_non_file(tmp_dir, dvc, file):
     (tmp_dir / file).mkdir()
+    dvcfile = Dvcfile(dvc, file)
     with pytest.raises(StageFileIsNotDvcFileError):
         assert dvcfile.stages.values()
 
@@ -83,3 +95,31 @@ def test_dump_stage(tmp_dir, dvc):
     assert (tmp_dir / PIPELINE_FILE).exists()
     assert (tmp_dir / PIPELINE_LOCK).exists()
     assert list(dvcfile.stages.values()) == [stage]
+
+
+@pytest.mark.parametrize("file", ["stage.dvc", "dvc.yaml"])
+def test_stage_load_file_exists_but_dvcignored(tmp_dir, dvc, scm, file):
+    (tmp_dir / file).write_text("")
+    (tmp_dir / ".dvcignore").write_text(file)
+
+    dvcfile = Dvcfile(dvc, file)
+    with pytest.raises(StageFileDoesNotExistError) as exc_info:
+        assert dvcfile.stages.values()
+
+    assert str(exc_info.value) == f"'{file}' is dvc-ignored"
+
+
+@pytest.mark.parametrize("file", ["foo.dvc", "dvc.yaml"])
+def test_try_loading_dvcfile_that_is_gitignored(tmp_dir, dvc, scm, file):
+    with open(tmp_dir / ".gitignore", "a+") as fd:
+        fd.write(file)
+
+    # create a file just to avoid other checks
+    (tmp_dir / file).write_text("")
+    scm._reset()
+
+    dvcfile = Dvcfile(dvc, file)
+    with pytest.raises(FileIsGitIgnored) as exc_info:
+        dvcfile._load()
+
+    assert str(exc_info.value) == f"'{file}' is git-ignored."