Optionally disallow duplicate labels #28394

TomAugspurger · 2019-09-11T22:13:09Z

This adds a property to NDFrame to disallow duplicate labels. This fixes a vexing issue with using pandas for ETL pipelines, where accidentally introducing duplicate labels can lead to confusing downstream behavior (e.g. NDFrame.__getitem__ not reducing dimensionality).

When set (via the construction with allow_duplicate_labels=False or afterward via .allows_duplicate_labels=False), the presence of duplicate labels causes a DuplicateLabelError exception to be raised:

In [3]: df = pd.DataFrame({"A": [1, 2]}, index=['a', 'a'], allow_duplicate_labels=False)
---------------------------------------------------------------------------
DuplicateLabelError                       Traceback (most recent call last)
<ipython-input-3-1c8833763dfc> in <module>
----> 1 df = pd.DataFrame({"A": [1, 2]}, index=['a', 'a'], allow_duplicate_labels=False)

~/sandbox/pandas/pandas/core/frame.py in __init__(self, data, index, columns, dtype, copy, allow_duplicate_labels)
    493
    494         NDFrame.__init__(
--> 495             self, mgr, fastpath=True, allow_duplicate_labels=allow_duplicate_labels
    496         )
    497

~/sandbox/pandas/pandas/core/generic.py in __init__(self, data, axes, copy, dtype, allow_duplicate_labels, fastpath)
    202
    203         if not self.allows_duplicate_labels:
--> 204             self._maybe_check_duplicate_labels()
    205
    206     def _init_mgr(self, mgr, axes=None, dtype=None, copy=False):

~/sandbox/pandas/pandas/core/generic.py in _maybe_check_duplicate_labels(self, force)
    240         if force or not self.allows_duplicate_labels:
    241             for ax in self.axes:
--> 242                 ax._maybe_check_unique()
    243
    244     @property

~/sandbox/pandas/pandas/core/indexes/base.py in _maybe_check_unique(self)
    540             # TODO: position, value, not too large.
    541             msg = "Index has duplicates."
--> 542             raise DuplicateLabelError(msg)
    543
    544     # --------------------------------------------------------------------

DuplicateLabelError: Index has duplicates.

This property is preserved through operations (using _metadata and __finalize__).

In [4]: df = pd.DataFrame(index=['a', 'A'], allow_duplicate_labels=False)

In [5]: df.rename(str.upper)
---------------------------------------------------------------------------
DuplicateLabelError                       Traceback (most recent call last)
<ipython-input-5-17c8fb0b7c7f> in <module>
----> 1 df.rename(str.upper)
...

~/sandbox/pandas/pandas/core/indexes/base.py in _maybe_check_unique(self)
    540             # TODO: position, value, not too large.
    541             msg = "Index has duplicates."
--> 542             raise DuplicateLabelError(msg)
    543
    544     # --------------------------------------------------------------------

DuplicateLabelError: Index has duplicates.

API design questions

Do we want to be positive or negative?

pd.Series(..., allow_duplicate_labels=True/False)
pd.Series(..., disallow_duplicate_labels=False/True)
pd.Series(..., require_unique_labels=False/True)

In my proposal, the argument is different from the property. Do we like that? The rational is the argument is making a statement about what to do, while the property is making a statement about what is allowed.

In [7]: s = pd.Series(allow_duplicate_labels=False)

In [8]: s.allows_duplicate_labels
Out[8]: False

I'd like a method-chainable way of saying "duplicates aren't allowed." Some options

# two methods.
s.disallow_duplicate().allow_duplicate()

I dislike that in combination with .allows_duplicates, since we'd have a property and a method that only differ by an s. Perhaps something like

s.update_metdata(allows_duplicates=False)

But do people know that "allows_duplicates" is considered metadata?

TODO:

many, many more tests
Propagate metadata in more places (groupby, rolling, etc.)
Backwards compatibility in __finalize__ (we may be changing when we call it, and what we pass)

Apologies for using the PR for discussion, but we needed a way to compare this to #28334.

pandas/core/generic.py

TomAugspurger · 2019-09-12T14:49:58Z

I think this is ready for an initial review, to decide if this direction / behavior is OK. I'd recommend starting with the new docs section for the behavior, and NDFrame.__finalize__ for the implementation.

cc @jorisvandenbossche @jreback @jbrockmendel @jschendel @datapythonista @ others I'm sure.

datapythonista · 2019-09-12T15:04:42Z

This looks very useful, thanks for the work on this. I can't think of a better API, so happy with your proposal.

Just wondering if in the future we may want allow_duplicate_values=False by default, but that's surely something to leave for later and keep backward-compatibility here.

TomAugspurger · 2019-09-12T15:16:38Z

Just wondering if in the future we may want allow_duplicate_values=False by default

Yeah I think that's a possibility. But yes let's leave it for later :)

jschendel · 2019-09-12T18:59:15Z

doc/source/user_guide/duplicates.rst

+or column labels. This may be a bit confusing at first. If you're familiar with
+SQL, you know that row labels are similar to a primary key on a table, and you
+would never want duplicates in a SQL table. But one of pandas' roles is to clean
+mess, real-world data before it goes to some downstream system. And real-world


mess --> messy?

jschendel · 2019-09-12T22:53:11Z

I'm +1 on this.

A couple passing comments:

I found it a little bit confusing that the constructor parameter is allow_duplicate_labels but the attribute name is slightly different in allows_duplicate_labels. My preference here would be to make these the same.
Is there any utility in making this configurable at the axis level, i.e. allow a dataframe to have duplicate index labels but not have duplicate column labels?
- I don't think this is terribly important but anticipate that we'll get a user request for that feature, so probably best to bring this up now.

One thing I noticed is that this doesn't appear to validate when you directly assign a new index:

In [2]: s = pd.Series(list('abc'), allow_duplicate_labels=False)

In [3]: s.index = [0, 0, 0]

In [4]: s
Out[4]: 
0    a
0    b
0    c
dtype: object

In [5]: s.allows_duplicate_labels
Out[5]: False

jbrockmendel · 2019-09-12T23:18:59Z

Once a kwarg is in the Series constructor it can be difficult to get it out (see: fastpath). Could we put this in stealth mode for a while with something like:

ser = pd.Series(call_constructor_like_always)
ser._allows_duplicate_labels = False

Not a strong opinion, just thinking out loud.

WillAyd · 2019-09-13T01:02:36Z

Just out of curiosity why do you think adding this as a keyword argument to the Series / DataFrame constructor is the right approach? Would that subsequently guard against doing something like ser.index = list("aaa") after construction?

TomAugspurger · 2019-09-13T18:42:06Z

I found it a little bit confusing that the constructor parameter is allow_duplicate_labels but the attribute name is slightly different in allows_duplicate_labels. My preference here would be to make these the same.

I share this concern. Do others? After using it for a few days I'm getting used to it, but still need to stop and think.

Is there any utility in making this configurable at the axis level, i.e. allow a dataframe to have duplicate index labels but not have duplicate column labels?

Yes. My backwards-compatible plan here is to have the setter allow True/False/index/columns.

* True : allows duplicates everywhere
* False : allows duplicates nowhere
* index : allows duplicates in the index, but not in the columns
* columns : allows duplicates in the columns, but not in the index.

Code-wise, it's not much effort to support. It just makes the logic a bit more complicated, so I'm holding off for now.

Once a kwarg is in the Series constructor it can be difficult to get it out (see: fastpath). Could we put this in stealth mode for a while with something like:

Do we anticipate needing to remove this? Certainly, it can be marked as experimental if there's trepidation.

Just out of curiosity why do you think adding this as a keyword argument to the Series / DataFrame constructor is the right approach?

Initially, I was thinking it'd be a property on Index. I think that's more natural, since it's making a statement about the Index. But users interact more with Series / DataFrame, so I think it's more ergonomic to place it there.

After implementing things, putting it on NDFrame is nice because we have NDFrame.__finalize__ If it were on the index, things would be a be trickier.

jbrockmendel · 2019-09-13T21:00:45Z

Do we anticipate needing to remove this? Certainly, it can be marked as experimental if there's trepidation.

I don't anticipate needing to remove this, but I do anticipate other conceptually similar flags being implemented. In that scenario, I don't like the idea of all of these becoming kwargs in the Series constructor.

This aids in the implementation of pandas-dev#28394. Over there, I'm having issues with using `NDFrame.__finalize__` to copy attributes, in part because getattribute on NDFrame is so complicated. This simplifies this because we only need to look in NDFrame.attrs, which is just a plain dictionary. Aside from the addition of a public NDFrame.attrs dictionary, there aren't any user-facing API changes.

commit 67a3263 Author: Tom Augspurger <tom.w.augspurger@gmail.com> Date: Fri Oct 18 08:05:04 2019 -0500 fixup name commit e6183cd Author: Tom Augspurger <tom.w.augspurger@gmail.com> Date: Fri Oct 18 07:05:33 2019 -0500 fixup Index.name commit d1826bb Author: Tom Augspurger <tom.w.augspurger@gmail.com> Date: Thu Oct 17 13:45:30 2019 -0500 REF: Store metadata in attrs dict This aids in the implementation of pandas-dev#28394. Over there, I'm having issues with using `NDFrame.__finalize__` to copy attributes, in part because getattribute on NDFrame is so complicated. This simplifies this because we only need to look in NDFrame.attrs, which is just a plain dictionary. Aside from the addition of a public NDFrame.attrs dictionary, there aren't any user-facing API changes.

TomAugspurger · 2019-10-23T21:55:13Z

I'm seeing some strange mypy issues with these changes. Lots of

pandas/tests/test_strings.py:206: error: List item 0 has incompatible type "Type[Series]"; expected "Type[PandasObject]"
pandas/tests/test_strings.py:240: error: List item 0 has incompatible type "Type[Series]"; expected "Type[PandasObject]"
...

L206 is @pytest.mark.parametrize("box", [Series, Index]).

In fact, I'm seeing these mypy errors with just the following diff from master

diff --git a/pandas/core/frame.py b/pandas/core/frame.py
index ef4e3e064d..7445951536 100644
--- a/pandas/core/frame.py
+++ b/pandas/core/frame.py
@@ -407,6 +407,7 @@ class DataFrame(NDFrame):
         columns: Optional[Axes] = None,
         dtype: Optional[Dtype] = None,
         copy: bool = False,
+        allow_duplicate_labels: bool = True,
     ):
         if data is None:
             data = {}
diff --git a/pandas/core/generic.py b/pandas/core/generic.py
index d59ce8db9b..64a38ca4eb 100644
--- a/pandas/core/generic.py
+++ b/pandas/core/generic.py
@@ -207,6 +207,7 @@ class NDFrame(PandasObject, SelectionMixin):
         dtype: Optional[Dtype] = None,
         attrs: Optional[Mapping[Optional[Hashable], Any]] = None,
         fastpath: bool = False,
+        allow_duplicate_labels: bool = True,
     ):
 
         if not fastpath:
diff --git a/pandas/core/series.py b/pandas/core/series.py
index 3e9d3d5c04..8d6beb42c4 100644
--- a/pandas/core/series.py
+++ b/pandas/core/series.py
@@ -203,7 +203,7 @@ class Series(base.IndexOpsMixin, generic.NDFrame):
     # Constructors
 
     def __init__(
-        self, data=None, index=None, dtype=None, name=None, copy=False, fastpath=False
+        self, data=None, index=None, dtype=None, name=None, copy=False, fastpath=False, allow_duplicate_labels=True,
     ):
 
         # we are called internally, so short-circuit

Any ideas what's going wrong (perhaps @simonjayhawkins or @WillAyd have a guess)?

WillAyd · 2019-10-23T22:02:24Z

What version of mypy / pytest are you using? Seems strange for that to pop up if that is your only diff

Working around a strange typing issue. See pandas-dev#28394 (comment) for more, but the types on these were being inferred incorrectly by mypy with just the addition of the `allows_duplicate_labels` kwarg.

TomAugspurger · 2019-10-24T14:01:23Z

pytest 5.2.1
mypy 0.720

Indeed, it's quite strange. Opened #29205 to add explicit types to the tests that were failing.

simonjayhawkins · 2019-10-24T15:24:43Z

on master

reveal_type([Series, Index]) gives Revealed type is 'builtins.list[builtins.type*]'

with the changes in this PR..

Revealed type is 'builtins.list[def (data: Any =, Any =, Any =, name: Any =, Any =, Any =, **Any) -> pandas.core.base.PandasObject]'

simonjayhawkins · 2019-10-24T15:46:46Z

same inferred type with mypy 0.730 and 0.740

jreback · 2020-01-01T17:21:43Z

@TomAugspurger status of this

TomAugspurger · 2020-01-01T17:39:08Z

Still interested in working on it, but not happening for for 1.0.

…

On Wed, Jan 1, 2020 at 11:21 AM Jeff Reback ***@***.***> wrote: @TomAugspurger <https://github.com/TomAugspurger> status of this — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#28394?email_source=notifications&email_token=AAKAOIWMFJXLSIRR6U3C5D3Q3TGKRA5CNFSM4IVZNIOKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEH5JALQ#issuecomment-570069038>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKAOISNENLHU2JZMPFPUC3Q3TGKRANCNFSM4IVZNIOA> .

TomAugspurger · 2020-08-25T11:01:21Z

Just the windows s3 failures, which I think that #35856 is tracking.

TomAugspurger · 2020-08-26T20:52:55Z

CI is green now.

jreback

nice updates

a few inline questions

can u ensure that this can survive a round trip pickle (flags)
worth separating out index and columns flags for duplicates? eg i almost always care about column duplicates but may want to allow row dupes

doc/source/user_guide/duplicates.rst

pandas/core/flags.py

pandas/core/frame.py

jreback

a couple of inline comments

can u ensure this survives a round trip pickle (flags)

worth having separate flags for index and columns ? eg disallow column but allow for rows i would say is common

the reason to do this now is that it would be very hard to change later

jreback · 2020-08-26T22:06:01Z

i think my comments got duped as internet not great

TomAugspurger · 2020-08-27T11:56:44Z

worth separating out index and columns flags for duplicates? eg i almost always care about column duplicates but may want to allow row dupes

Possibly. The current system is future-compatible with that though, allows_duplicate_labels="index" / "columns"

TomAugspurger · 2020-08-27T12:06:37Z

Added the pickle test.

TomAugspurger · 2020-08-31T12:50:55Z

All green with the pickle test now.

I've also added tests that assert_eq correctly compares .flags. I think we would (by default) consider tm.assert_frame_equal(a, b) to be False when the flags don't match.

jreback

couple of minor comments and questions. looks good

doc/source/user_guide/duplicates.rst

doc/source/whatsnew/v1.2.0.rst

pandas/core/generic.py

pandas/core/indexes/base.py

jreback · 2020-08-31T23:57:26Z

pandas/core/indexes/base.py

@@ -483,6 +483,26 @@ def _simple_new(cls, values, name: Label = None):
    def _constructor(self):
        return type(self)

+    def _maybe_check_unique(self):
+        if not self.is_unique:
+            # TODO: position, value, not too large.


yeah maybe an arg to show the first 10 duplicates

It's truncated, following our usual repr's settings. I'll remove the todo

In [3]: s = pd.Series(1, index=[0] * 200).set_flags(allows_duplicate_labels=False) --------------------------------------------------------------------------- DuplicateLabelError Traceback (most recent call last) <ipython-input-3-ece63ed32110> in <module> ----> 1 s = pd.Series(1, index=[0] * 200).set_flags(allows_duplicate_labels=False) ~/sandbox/pandas/pandas/core/generic.py in set_flags(self, copy, allows_duplicate_labels) 349 df = self.copy(deep=copy) 350 if allows_duplicate_labels is not None: --> 351 df.flags["allows_duplicate_labels"] = allows_duplicate_labels 352 return df 353 ~/sandbox/pandas/pandas/core/flags.py in __setitem__(self, key, value) 103 if key not in self._keys: 104 raise ValueError(f"Unknown flag {key}. Must be one of {self._keys}") --> 105 setattr(self, key, value) 106 107 def __repr__(self): ~/sandbox/pandas/pandas/core/flags.py in allows_duplicate_labels(self, value) 90 if not value: 91 for ax in obj.axes: ---> 92 ax._maybe_check_unique() 93 94 self._allows_duplicate_labels = value ~/sandbox/pandas/pandas/core/indexes/base.py in _maybe_check_unique(self) 491 msg += "\n{}".format(duplicates) 492 --> 493 raise DuplicateLabelError(msg) 494 495 def _format_duplicate_message(self): DuplicateLabelError: Index has duplicates. positions label 0 [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,...

In [10]: s = pd.Series(1, index=np.arange(100).repeat(10)).set_flags(allows_duplicate_labels=False) --------------------------------------------------------------------------- DuplicateLabelError Traceback (most recent call last) <ipython-input-10-452fca37a730> in <module> ----> 1 s = pd.Series(1, index=np.arange(100).repeat(10)).set_flags(allows_duplicate_labels=False) ~/sandbox/pandas/pandas/core/generic.py in set_flags(self, copy, allows_duplicate_labels) 349 df = self.copy(deep=copy) 350 if allows_duplicate_labels is not None: --> 351 df.flags["allows_duplicate_labels"] = allows_duplicate_labels 352 return df 353 ~/sandbox/pandas/pandas/core/flags.py in __setitem__(self, key, value) 103 if key not in self._keys: 104 raise ValueError(f"Unknown flag {key}. Must be one of {self._keys}") --> 105 setattr(self, key, value) 106 107 def __repr__(self): ~/sandbox/pandas/pandas/core/flags.py in allows_duplicate_labels(self, value) 90 if not value: 91 for ax in obj.axes: ---> 92 ax._maybe_check_unique() 93 94 self._allows_duplicate_labels = value ~/sandbox/pandas/pandas/core/indexes/base.py in _maybe_check_unique(self) 491 msg += "\n{}".format(duplicates) 492 --> 493 raise DuplicateLabelError(msg) 494 495 def _format_duplicate_message(self): DuplicateLabelError: Index has duplicates. positions label 0 [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] 1 [10, 11, 12, 13, 14, 15, 16, 17, 18, 19] 2 [20, 21, 22, 23, 24, 25, 26, 27, 28, 29] 3 [30, 31, 32, 33, 34, 35, 36, 37, 38, 39] 4 [40, 41, 42, 43, 44, 45, 46, 47, 48, 49] ... ... 95 [950, 951, 952, 953, 954, 955, 956, 957, 958, ... 96 [960, 961, 962, 963, 964, 965, 966, 967, 968, ... 97 [970, 971, 972, 973, 974, 975, 976, 977, 978, ... 98 [980, 981, 982, 983, 984, 985, 986, 987, 988, ... 99 [990, 991, 992, 993, 994, 995, 996, 997, 998, ... [100 rows x 1 columns]

jreback

lgtm. thanks @TomAugspurger

TomAugspurger · 2020-09-03T11:26:22Z

Thanks!

…

On Wed, Sep 2, 2020 at 10:00 PM Jeff Reback ***@***.***> wrote: Merged #28394 <#28394> into master. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#28394 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKAOIV6VNVMWLXV2KW75ZLSD4BD5ANCNFSM4IVZNIOA> .

TomAugspurger mentioned this pull request Sep 11, 2019

REF/ENH: Refactor NDFrame finalization #28334

Closed

3 tasks

TomAugspurger commented Sep 11, 2019

View reviewed changes

pandas/core/generic.py Outdated Show resolved Hide resolved

jschendel reviewed Sep 12, 2019

View reviewed changes

TomAugspurger added this to the 1.0 milestone Sep 13, 2019

TomAugspurger added metadata _metadata, .attrs API Design labels Sep 13, 2019

TomAugspurger mentioned this pull request Oct 17, 2019

REF: Store metadata in an attrs dict #29062

Merged

TomAugspurger force-pushed the unique-index branch from 9cc17b7 to 05e238d Compare October 21, 2019 20:55

TomAugspurger force-pushed the unique-index branch from 5b7662f to d771b80 Compare October 22, 2019 16:16

TomAugspurger mentioned this pull request Oct 24, 2019

TYPING: Added types for some tests #29205

Closed

TomAugspurger mentioned this pull request Nov 6, 2019

DataFrame.groupby doesn't preserve _metadata #29442

Closed

jreback removed this from the 1.0 milestone Jan 1, 2020

TomAugspurger added 2 commits August 24, 2020 17:33

lint

74a4eb8

lint

50042e1

Merge remote-tracking branch 'upstream/master' into unique-index

6a1560f

jreback requested changes Aug 26, 2020

View reviewed changes

doc/source/user_guide/duplicates.rst Outdated Show resolved Hide resolved

pandas/core/flags.py Show resolved Hide resolved

pandas/core/frame.py Show resolved Hide resolved

jreback requested changes Aug 26, 2020

View reviewed changes

TomAugspurger added 3 commits August 27, 2020 06:58

Merge remote-tracking branch 'upstream/master' into unique-index

f61118d

Pickle

a336027

Doc

60c853c

TomAugspurger added 4 commits August 27, 2020 07:18

test flags

cb5d4f2

pickle

98af7b5

remove debug

3a50741

Merge remote-tracking branch 'upstream/master' into unique-index

675e49d

jreback requested changes Aug 31, 2020

View reviewed changes

TomAugspurger added 2 commits September 1, 2020 08:21

Merge remote-tracking branch 'upstream/master' into unique-index

e88f7e2

fixups

bf23fda

jreback approved these changes Sep 3, 2020

View reviewed changes

jreback merged commit 76eb314 into pandas-dev:master Sep 3, 2020

TomAugspurger deleted the unique-index branch September 8, 2020 16:44

TomAugspurger mentioned this pull request Sep 21, 2020

API: Add NDFrame property to disallow duplicates #27108

Closed

kesmit13 pushed a commit to kesmit13/pandas that referenced this pull request Nov 2, 2020

Optionally disallow duplicate labels (pandas-dev#28394)

4b03ee6

jorisvandenbossche mentioned this pull request Mar 27, 2023

DEPR: flags #52165

Open

joelostblom mentioned this pull request May 13, 2023

ENH: Disallow duplicate column names everywhere by default #53217

Open

3 tasks

Dr-Irv mentioned this pull request Dec 27, 2023

DEPR: flags #52153

Closed

5 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Optionally disallow duplicate labels #28394

Optionally disallow duplicate labels #28394

TomAugspurger commented Sep 11, 2019 •

edited

Loading

TomAugspurger commented Sep 12, 2019

datapythonista commented Sep 12, 2019

TomAugspurger commented Sep 12, 2019

jschendel Sep 12, 2019

jschendel commented Sep 12, 2019

jbrockmendel commented Sep 12, 2019

WillAyd commented Sep 13, 2019

TomAugspurger commented Sep 13, 2019

jbrockmendel commented Sep 13, 2019

TomAugspurger commented Oct 23, 2019

WillAyd commented Oct 23, 2019

TomAugspurger commented Oct 24, 2019

simonjayhawkins commented Oct 24, 2019

simonjayhawkins commented Oct 24, 2019

jreback commented Jan 1, 2020

TomAugspurger commented Jan 1, 2020 via email

TomAugspurger commented Aug 25, 2020

TomAugspurger commented Aug 26, 2020

jreback left a comment

jreback left a comment

jreback commented Aug 26, 2020

TomAugspurger commented Aug 27, 2020

TomAugspurger commented Aug 27, 2020

TomAugspurger commented Aug 31, 2020 •

edited

Loading

jreback left a comment

jreback Aug 31, 2020

TomAugspurger Sep 1, 2020

jreback left a comment

TomAugspurger commented Sep 3, 2020 via email

Optionally disallow duplicate labels #28394

Optionally disallow duplicate labels #28394

Conversation

TomAugspurger commented Sep 11, 2019 • edited Loading

API design questions

TomAugspurger commented Sep 12, 2019

datapythonista commented Sep 12, 2019

TomAugspurger commented Sep 12, 2019

jschendel Sep 12, 2019

Choose a reason for hiding this comment

jschendel commented Sep 12, 2019

jbrockmendel commented Sep 12, 2019

WillAyd commented Sep 13, 2019

TomAugspurger commented Sep 13, 2019

jbrockmendel commented Sep 13, 2019

TomAugspurger commented Oct 23, 2019

WillAyd commented Oct 23, 2019

TomAugspurger commented Oct 24, 2019

simonjayhawkins commented Oct 24, 2019

simonjayhawkins commented Oct 24, 2019

jreback commented Jan 1, 2020

TomAugspurger commented Jan 1, 2020 via email

TomAugspurger commented Aug 25, 2020

TomAugspurger commented Aug 26, 2020

jreback left a comment

Choose a reason for hiding this comment

jreback left a comment

Choose a reason for hiding this comment

jreback commented Aug 26, 2020

TomAugspurger commented Aug 27, 2020

TomAugspurger commented Aug 27, 2020

TomAugspurger commented Aug 31, 2020 • edited Loading

jreback left a comment

Choose a reason for hiding this comment

jreback Aug 31, 2020

Choose a reason for hiding this comment

TomAugspurger Sep 1, 2020

Choose a reason for hiding this comment

jreback left a comment

Choose a reason for hiding this comment

TomAugspurger commented Sep 3, 2020 via email

TomAugspurger commented Sep 11, 2019 •

edited

Loading

TomAugspurger commented Aug 31, 2020 •

edited

Loading