Cache hash codes #426

gabbard · 2018-08-19T01:10:26Z

Pull Request Check List

This is just a reminder about the most common mistakes. Please make sure that you tick all appropriate boxes. But please read our contribution guide at least once, it will save you unnecessary review cycles!

Added tests for changed code.
New features have been added to our Hypothesis testing strategy. ** n/a,I think**
Changes or additions to public APIs are reflected in our type stubs (files ending in .pyi).
Updated documentation for changed code.
Documentation in .rst files is written using semantic newlines.
Changed/added classes/methods/functions have appropriate versionadded, versionchanged, or deprecated directives.
Changes (and possible deprecations) have news fragments in changelog.d.

Description

This adds a cache_hash flag to @attrs which causes the hash code to be computed once and stored on the object. Setting this flag to true is allowed only if attrs is set to generate both the __hash__ and __init__ methods for a class.

This closes issues #423.

Currently all existing tests pass but no cache_hash tests have yet been added.

gabbard · 2018-08-19T02:44:39Z

@hynek : I believe this is now ready for review

hynek

For a PR of this size this looks quite excellent! I’ve added a bunch of nits, but this should be mergeable very soon!

src/attr/_make.py

+                raise TypeError(
+                    "Invalid value for cache_hash.  To use hash caching,"
+                    " hashing must be either explicitly or implicitly "
+                    "enabled"


src/attr/_make.py

+                raise TypeError(
+                    "Invalid value for cache_hash.  To use hash caching,"
+                    " hashing must be either explicitly or implicitly "
+                    "enabled"


src/attr/_make.py

+            if cache_hash:
+                raise TypeError(
+                    "Invalid value for cache_hash.  To use hash caching,"
+                    " init must be True"


tests/test_dunders.py

@@ -283,8 +293,34 @@ def test_enforces_type(self):

        assert exc_args == e.value.args

-    @given(booleans())
-    def test_hash_attribute(self, slots):
+    def test_enforce_no_cache_hash_without_hash(self):


changelog.d/425.change.rst

@@ -0,0 +1 @@
+Added ``cache_hash`` option which causes the hash code to be computed once and stored on the object.


docs/hashing.rst

+
+Some objects have hash codes which are expensive to compute.
+If such objects are to be stored in hash-based collections, it can be useful to compute the hash codes only once and then store the result on the object to make future hash code requests fast.
+To enable caching of hash codes, specify ``cache_hash=True``.


docs/hashing.rst

+To enable caching of hash codes, specify ``cache_hash=True``.
+This may only be done if ``attrs`` is already generating a hash function for the object.
+If the hash code is cached, no field involved in hash code computation may be mutated after construction.
+It is strongly recommended that classes with cached hashcodes be ``frozen``


docs/init.rst

@@ -348,6 +348,7 @@ If you need to set attributes on a frozen class, you'll have to resort to the :r
   >>> Frozen(1)
   Frozen(x=1, y=2)

+Note that you may not access the hash code of the object in ``__attrs_post__init__`` if ``cache_hash=True``


src/attr/_make.py

+
+    if cache_hash:
+        method_lines.append(tab + "if self.%s is None:" % _hash_cache_field)
+        # we use __setattr__ all the time so we don't need to special-case for


tests/test_dunders.py

+            )
+        assert exc_args == e.value.args
+
+    def test_enforce_no_cached_hash_without_init(self):


gabbard · 2018-08-19T18:04:44Z

@hynek : Ready for re-review. For all the trivial changes (periods, adding phrases to docs, docstrings, etc.) I made the suggested changes. In the two more complex cases, I commented on your comments to highlight my changes.

hynek · 2018-08-20T04:47:00Z

Great work, thank you!

gabbard · 2018-08-20T16:07:53Z

@hynek : no problem, thanks for reviewing and merging!

Do you happen to know when the next release is likely to be? My team would really like to use this feature and the new kw-only args support.

hynek · 2018-08-21T03:57:13Z

Yeah it’s overdue, isn’t it. I want to fix #429 and #427 since both look like bugs and then release. So I hope this week if I find people to review my fixes.

gabbard · 2018-08-21T14:31:55Z

@hynek : Great, thank you!

Ryan Gabbard added 10 commits August 17, 2018 14:38

First stab at implementing hashcode caching (python-attrs#423)

4066c60

Currently all existing tests pass but no cache_hash tests have yet been added.

Existing hash tests now pass on cache_hash classes

19b9248

Add towncrier change log

c5f20be

Add documentation for cache_hash

5b1b45e

Fixes bug with check that init=True if cache_hash=True

286ba00

Fix long lines

f6452aa

Fix documentation issues

a081bbd

Add test for cache_hash requiring init

675ff99

Improve test coverage

53ac952

Remove now unnecessary 'pass'

38c4d08

hynek requested changes Aug 19, 2018

View reviewed changes

Ryan Gabbard added 5 commits August 19, 2018 13:16

Add periods to the end of exception strings

7d8ed9a

Add test docstrings for cache_hash tests

3223045

Clarify documentation of cache_hash

5bdd739

Recommend that hashable classes be frozen

c9f5669

Fix test references for exception messages

f81d319

hynek approved these changes Aug 20, 2018

View reviewed changes

hynek merged commit 3e0ecbd into python-attrs:master Aug 20, 2018

gabbard deleted the 423-cache-hash branch August 20, 2018 16:08

mattsb42-aws mentioned this pull request Aug 22, 2018

Some new validators #425

Merged

6 tasks

gabbard mentioned this pull request Oct 22, 2018

Caching hashcodes #423

Closed

gabbard mentioned this pull request Jan 11, 2019

cache_hash can give the wrong hash code for deserialized objects #482

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Cache hash codes #426

Cache hash codes #426

gabbard commented Aug 19, 2018 •

edited

Loading

gabbard commented Aug 19, 2018

hynek left a comment

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

gabbard commented Aug 19, 2018

hynek commented Aug 20, 2018

gabbard commented Aug 20, 2018

hynek commented Aug 21, 2018

gabbard commented Aug 21, 2018

		@@ -0,0 +1 @@
		Added ``cache_hash`` option which causes the hash code to be computed once and stored on the object.

Cache hash codes #426

Cache hash codes #426

Conversation

gabbard commented Aug 19, 2018 • edited Loading

Pull Request Check List

Description

gabbard commented Aug 19, 2018

hynek left a comment

Choose a reason for hiding this comment

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

This comment was marked as spam.

gabbard commented Aug 19, 2018

hynek commented Aug 20, 2018

gabbard commented Aug 20, 2018

hynek commented Aug 21, 2018

gabbard commented Aug 21, 2018

gabbard commented Aug 19, 2018 •

edited

Loading