A common test pattern is to create a Commit whose tree is a SourceTree
defined by a Source that is defined by a SourceSpec.
Rather than instantiating this in a complex nested call of constructors
every time, this new Commit.from_spec constructor classmethod does it
all at once, distributing the correct keyword arguments to the correct
constructors.
This allows much more concise test input data both for existing tests
(refactored) and for tests that will be added later in this branch.
It's clearer if validation_repo is computed only immediately before the
assertions are made. This allows tests to better follow the pattern
(setup, execute, verify); computing validation_repo is only necessary,
by definition, in the verify step.
Allow Repo() and Repo.copy() to take default "nothing" arguments, and
make the terminology consistent with behaviour ("update" matches
dict.update; "list" is implicit as a Python sequence). Clean up all
calls to Repo() and Repo.copy().
Use "input_repo" to describe the test parameter that supplies a input
repository to the behaviour under test.
Use "validation_repo" to describe a repository that is to be compared
against to verify an expected result in the case that it isn't exactly
the expected repository.
Use "validation_repo_delta" to describe how to derive validation_repo
from input_repo.
Use "validation_repo_expected_identical_refs" to describe a subset of
references inside validation_repo that are expected to be identical
between validation_repo and actual results.
This change is not intended to make any functional change to the tests.
Docstrings are added where they were previously missing.
It seems to be far more common that tests use non-native fake source
packages, so make this the default. This simplifies the call to the
SourceSpec constructor in many tests.
An empty tree is a common pattern in tests, so permit it to not be
specified and default to an empty tree if it is not. This makes many
tests more concise.
Placeholder is used everywhere in this file, so import it into our
namespace directly to tidy up all the references to it. This makes the
test input data easier to read.
