ENH: MultiIndex.from_frame #23141

ms7463 · 2018-10-13T22:20:38Z

closes Feature Request: pd.MultiIndex.from_frame(). Complement to pd.MultiIndex.to_frame(). #22420
tests added / passed
passes git diff upstream/master -u -- "*.py" | flake8 --diff
whatsnew entry

This pull request is to add the from_frame method of creating multiindexes. Along with this feature the helper method "squeeze" has been added for squeezing single level multiindexes to a standard index (analogous to df.squeeze).

Additionally the to_frame method was updated to guarantee that the order of the labels is preserved when converting a multiindex to a dataframe. Currently this cannot be guaranteed in Python 2.7 due to the use of a dictionary for creating the frame. With this change from_frame and to_frame are perfectly complementary.

…ndex

…ame method

pep8speaks · 2018-10-13T22:20:41Z

Hello @ArtinSarraf! Thanks for updating the PR.

There are no PEP8 issues in the file pandas/core/indexes/multi.py !
There are no PEP8 issues in the file pandas/tests/indexes/multi/test_constructor.py !
There are no PEP8 issues in the file pandas/tests/indexes/multi/test_conversion.py !

Comment last updated on November 19, 2018 at 04:25 Hours UTC

jreback · 2018-10-13T22:33:59Z

huh? pls show a real usecase for this

in the future pls open an issue for discussion first

ms7463 · 2018-10-13T22:36:32Z

Sorry should have tagged this issue opened ~2 months ago. The issue contains an example as well.

#22420

ms7463 · 2018-10-13T23:04:52Z

Here are some more concrete examples of how these can be used as well.

Adding metadata.

df = pd.DataFrame(np.random.randn(10, 6), columns=pd.MultiIndex.from_tuples([('NY', 'KJFK'), ('NY', 'KLGA'), ('CA', 'KLAX'), ('CA', 'KBUR'), ('AB', 'CYEG'), ('AB', 'CYYC')], names=['state', 'ICAO']))
country_state_mapping = pd.DataFrame([['US', 'NY'], ['US', 'CA'], ['CDN', 'AB'], ['CDN', 'ON']], columns=['country', 'state'])
meta = df.columns.to_frame(index=False)
new_meta = meta.merge(country_state_mapping, on='state', how='left')[['country', 'state', 'ICAO']]
df.columns = pd.MultiIndex.from_frame(new_meta)  # this can only work if to_frame guarantees label ordering, which it currently does not in 2.7, but does with this PR.

Filtering

# continuing from previous
meta = df.columns.to_frame(index=False)
new_meta = meta.loc[meta.state.isin(['CA', 'AB'])]
df = df.reindex(columns=pd.MultiIndex.from_frame(new_meta))

The examples here are extremely simplified, but reflect how my coworkers and I use pandas for working with multiindices.

codecov · 2018-10-13T23:37:23Z

Codecov Report

Merging #23141 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master   #23141      +/-   ##
==========================================
+ Coverage    92.2%    92.2%   +<.01%     
==========================================
  Files         162      162              
  Lines       51700    51709       +9     
==========================================
+ Hits        47670    47679       +9     
  Misses       4030     4030

Flag	Coverage Δ
#multiple	`90.6% <100%> (ø)`	⬆️
#single	`43.02% <40%> (ø)`	⬆️

Impacted Files	Coverage Δ
pandas/core/indexes/multi.py	`95.58% <100%> (+0.03%)`	⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fc64ca8...9159b2d. Read the comment docs.

jreback · 2018-10-14T13:11:24Z

you can do the same thing if everything is in long form.

In [93]: (pd.merge(df.unstack().reset_index(), country_state_mapping, on='state', how='left')
    ...:    .set_index(['country', 'state', 'ICAO', 'level_2'])
    ...:    .sort_index()
    ...:    .unstack(level=[0,1,2])
    ...:    .sort_index(axis=1)
    ...: )
Out[93]: 
                0                                                  
country       CDN                  US                              
state          AB                  CA                  NY          
ICAO         CYEG      CYYC      KBUR      KLAX      KJFK      KLGA
level_2                                                            
0        0.368735  0.428429 -1.173573  0.256197  1.792880  0.414737
1        0.993952 -0.313126 -1.814687 -0.025160 -0.775963 -0.665919
2       -0.635185  2.063095  0.063650 -0.939815  0.219432  1.489314
3       -0.156358  0.201501  0.344305 -1.562817 -0.103385  0.482686
4        0.551272 -1.628747 -1.387776 -1.400052 -1.493635  2.323896
5       -1.487852  0.466714  1.195675  0.006298  0.277463  1.170253
6       -0.453253  1.850580  0.240695  1.375471  0.144580  0.012171
7       -0.079981 -1.139398  0.855885  2.687475 -1.374322  0.543441
8       -1.432287  0.100697 -0.554310 -0.306952  1.011671 -0.723025
9       -0.235468  0.893887  0.638946  0.916163  0.149475  1.121286

What you are proposing is wide from manipulation, which though a nice idea, and in this particular case makes sense. Is not generaly purpose, nor easy for people to do (and requires sorting guarantees).

So why should this be added?

ms7463 · 2018-10-14T17:52:17Z

One of the nice features of multiindexes is that they act as meta-dataframes for the associated numeric data, allowing a frame to be much more concise and memory efficient. Long form manipulations require multiple pivots (which can be time consuming) and can require significant extra memory. The relationship between multiindexes and a standard dataframe can also be demonstrated through this following example. One can substitute a multiindex (and I have seen this done before) with a separate metaframe like so:

df = pd.DataFrame(np.random.randn(10, 6))
meta = pd.DataFrame([('NY', 'KJFK'), ('NY', 'KLGA'), ('CA', 'KLAX'), ('CA', 'KBUR'), ('AB', 'CYEG'), ('AB', 'CYYC')], columns=['state', 'ICAO'])
df.loc[:, meta.loc[meta.state.isin(['NY', 'AB'])].index]  # equivalent to above filtering example

I believe that this, along with the fact that to_frame already exists shows that from_frame is a reasonable and logical complementary method to be added as a way to instantiate multiindexes.

Regarding the sorting guarantees of to_frame, I think that this is something that any user would reasonably expect, considering the similarities between a mi and a df. Additionally, not only does it keep manipulations, that might involve using it, consistent, but all other related casting methods (e.g. tolist) ensure the sorting of the resultant type matches that of the original multiindex. At the very least I think adding a parameter to to_frame for preserving the original sorting would make sense. It would be very easy to add to the existing implementation.

jreback

so need some discussions

and of course would need much testing - having some examples would acceptance possibikity

ms7463 · 2018-10-14T22:01:25Z

I would be happy to contribute to any further discussions. Is there a more appropriate forum for that?

Also I will add tests to the PR this week.

Are there any types of examples in particular that would help move along the addition of this feature?

Thanks.

jreback · 2018-10-15T12:43:34Z

Are there any types of examples in particular that would help move along the addition of this feature?

some usecases, and showing how its possible now, but how the 'new' method makes it easeir.

ms7463 · 2018-10-16T02:57:34Z

Here is a demo to show the benefits of the from_frame method. Thanks for taking the time to look.
Some tests have been added as well.

# for convenience of this demo
pd.MultiIndex.from_frame = staticmethod(lambda df: pd.MultiIndex.from_tuples(list(df.values), names=list(df)))

Setup df

# memory usage of below frame ~82000000
#    multiindexes this large are not uncommon when dealing with things such as weather data for many cities
#    with their associated meta location info (i.e. country, state, county, etc.)
df = df = pd.DataFrame(
    np.random.randn(10000, 1024), 
    index=pd.date_range('19910905', periods=10000, name='date'), 
    columns=pd.MultiIndex.from_product(
        [list('abcd'), list('efgh'), list('ijkl'), list('mnop'), list('qrst')], 
        names=['L1', 'L2', 'L3', 'L4', 'L5']
    )
)

meta_mapping = pd.DataFrame([['x', 'a'], ['x', 'b'], ['y', 'c'], ['z', 'd']], columns=['L0', 'L1'])

Adding meta data

# Long Form
# peak memory usage ~573440080  --  ~7x larger than original df
# execution time: ~13.7 secs

(
    df.unstack().reset_index().merge(meta_mapping, on='L1', how='left')
    .set_index(['L0', 'L1', 'L2', 'L3', 'L4', 'L5', 'date'])[0].unstack([0, 1, 2, 3, 4, 5])
)


# Wide Form
# peak memory usage ~82000000  --  no difference from original df
# execution time: ~0.271 sec  --  ~50x faster than long form
# cleaner and does not require explicitly writing out positional arguments or axis 0 names
# requires guaranteed ordering of to_frame - easy to guarantee/implement

(
    df.T.set_index(pd.MultiIndex.from_frame(
        df.columns.to_frame(index=False)
        .merge(meta_mapping, on='L1', how='left')[['L0', 'L1', 'L2', 'L3', 'L4', 'L5']]
    )).T
)

Quick Prototyping

Use this df for the example below

# Often find it easier to prototype meta frames in excel. I then need to use that frame as a multiindex.
# Assume the below snippet was copied from excel rather than pyperclip
pyperclip.copy('''
country   state   county        city       
USA       ct      fairfield     Greenwich  
USA       ct      fairfield     Stamford   
USA       ny      westchester   Bronxville 
USA       ny      bronx         Bronx      
USA       nj      monmouth      Rumson     
USA       nj      monmouth      Middletown 
USA       nj      ocean         Ocean      
Canada    on      york          Toronto    
''')
meta = pd.read_clipboard()
df = pd.DataFrame(np.random.randn(10, 8), columns=pd.MultiIndex.from_frame(meta))

# As far as I know there is no good and quick way to do this currently. (aside from the from_tuples method I showed above)

Formatting MultiIndex

Use this formatted df for remainder of demo

# With from_frame
meta = df.columns.to_frame(index=False)
meta = meta.assign(state=meta.state.str.upper(), county=meta.county.str.title())
df.columns = pd.MultiIndex.from_frame(meta)

# Without from_frame
# More verbose, not as clear and requires the use of positional levels (as far as I know there is no complementary set_level_values method for get_level_values, which would also be nice)
df.columns = df.columns.set_levels(df.columns.levels[1].str.upper(), level=1).set_levels(df.columns.levels[2].str.title(), level=2)

Complex Filtering/Slicing

meta = df.columns.to_frame(index=False)  # used for all from_frame examples below

Keep only counties with > 1 city

# With from_frame
df.reindex(columns=pd.MultiIndex.from_frame(meta.groupby('county').filter(lambda df: len(df)>1)))

# Long Form - same disadvantages as stated above
(
    df.unstack().reset_index().groupby('county').filter(lambda dfx: len(dfx.city.unique())>1)
    .set_index(['country', 'state', 'county', 'city', 'level_4']).unstack([0,1,2,3])[0]
)

Admittedly the 2 remaining examples below can be done without from_frame as well, like so:

# Requires to_frame to be ordered
mask = (meta...) & (meta...) | (meta...)
df.loc[:, mask]

Get CT + NJ without Greenwich

# With from_frame
filtered_df = meta.loc[(meta.state.isin(['CT', 'NJ'])) & (meta.city != 'Greenwich')]
df.reindex(columns=pd.MultiIndex.from_frame(filtered_df))

# Without from_frame
# While this is more concise, it requires two different chained methods, while the from_frame
# method simplifies the filter to a standard dataframe filter
df.loc[:, pd.IndexSlice['USA', ['CT', 'NJ'], :]].drop('Greenwich', axis=1, level='city')

Get all cities that don't start with "Bron"

# With from_frame
filtered_df = meta.loc[~meta.city.str.startswith('Bron')]
df.reindex(columns=pd.MultiIndex.from_frame(filtered_df))

# Without from_frame
# Not sure of a good way without going to long form

Custom Sorting

This is nice for formatting report output

Given the following custom ordering

order = {'state': ['NJ', 'CT', 'NY', 'ON'], 'country': ['USA', 'Canada']}

# With from_frame
def custom_sort(df, order):  # would be nice to see something like this as a df method (with some edge case handling)
    df = df.copy()
    orig_dtypes = df.dtypes.to_dict()
    for col in df:
        if col in order:
            df.loc[:, col] = pd.Categorical(df.loc[:, col], order[col])
    df = df.sort_values(by=df.columns.tolist())
    for col in df:
        df[col] = df[col].astype(orig_dtypes[col])
    return df

meta = df.columns.to_frame(index=False)
df.reindex(columns=pd.MultiIndex.from_frame(custom_sort(meta, order)))

# Without from_frame
# However, this requires explicitly listing out all the columns
# Not ideal/practical if you have many more items
df.iloc[:, [5,4,6,0,1,3,2,7]]

ms7463 · 2018-10-16T14:01:49Z

Is there a way to rerun the Azure pipeline tests? The tests only failed for “Windows py36_np14” environment and the failed tests don’t seem to be related to my commits at all. I’m guessing it was a blip in the testing framework.

ms7463 · 2018-10-20T14:07:54Z

@jreback

Please see additions to demo (Formatting and Custom Sorting):
#23141 (comment)

Also any ideas on the tests? I don't see anything in the logs that point to any changes I've made.

ms7463 · 2018-10-27T16:42:00Z

@jreback. Any thoughts on the examples provided above?

jreback · 2018-10-28T03:29:54Z

@pandas-dev/pandas-core if anyone has thoughts here

datapythonista

Some issues with the docstrings. Also would be nice to add See Also sections. If you can take a look at the documentation for the docstrings that would be great.

datapythonista · 2018-10-28T03:43:16Z

pandas/core/indexes/multi.py

+        Parameters
+        ----------
+        df : pd.DataFrame
+            DataFrame to be converted to MultiIndex


Can you finish with a period.

datapythonista · 2018-10-28T03:43:32Z

pandas/core/indexes/multi.py

+        ----------
+        df : pd.DataFrame
+            DataFrame to be converted to MultiIndex
+        squeeze : bool


Can you add the default.

datapythonista · 2018-10-28T03:45:17Z

pandas/core/indexes/multi.py

+    @classmethod
+    def from_frame(cls, df, squeeze=True):
+        """
+        Make a MultiIndex from a dataframe


Finish with period. Run ./scripts/validate_docstrings.py pandas.MultiIndex.from_frame to make sure the docstring follows all the standards we validate.

Use DataFrame instead of dataframe.

datapythonista · 2018-10-28T03:46:01Z

pandas/core/indexes/multi.py

+
+        Returns
+        -------
+        index : MultiIndex


no need for giving a name, just leave the type. Add a description in the next line (indented).

datapythonista · 2018-10-28T03:48:28Z

pandas/core/indexes/multi.py

+        --------
+        >>> df = pd.DataFrame([[0, u'green'], [0, u'purple'], [1, u'green'],
+                               [1, u'purple'], [2, u'green'], [2, u'purple']],
+                              columns=[u'number', u'color'])


this is missing the ... for multine commands. Also show the content of df after creating it.

datapythonista · 2018-10-28T03:50:02Z

pandas/core/indexes/multi.py

+        """
+        Squeeze a single level multiindex to be a regular Index instane. If
+        the MultiIndex is more than a single level, return a copy of the
+        MultiIndex.


Can you run validate_docstrings.py for this docstring, and also read the contributing docstring documentation too. There are several issues.

…h the Pandas docstring guide

ms7463 · 2018-10-28T17:20:19Z

@datapythonista. Changes have been made and pushed. I've also added from_frame to the "See Also" of the other 3 contructor methods. Thanks.

…af/pandas into enhancement/from_frame

jreback

some doc comments. can you update advanced.rst (see where we use .from_arrays and add an example using .from_frame). ping on green.

jreback · 2018-12-04T03:53:54Z

pandas/tests/indexes/multi/test_constructor.py

+@pytest.mark.parametrize('names_in,names_out', [
+    (None, [('L1', 'x'), ('L2', 'y')]),
+    (['x', 'y'], ['x', 'y']),
+    ('bad_input', ValueError("Names should be list-like for a MultiIndex")),


why don't you split this to 2 tests, with 1 the working cases, and 1 the error cases, easier to read

jreback · 2018-12-04T03:56:04Z

looks pretty good to me, just some doc comments remain. @toobaz @datapythonista can you have a look and comment or approve.

jreback

small doc-comments. ping on green.

doc/source/advanced.rst

jreback · 2018-12-06T01:24:53Z

doc/source/whatsnew/v0.24.0.rst

@@ -378,6 +378,7 @@ Backwards incompatible API changes
 - Passing scalar values to :class:`DatetimeIndex` or :class:`TimedeltaIndex` will now raise ``TypeError`` instead of ``ValueError`` (:issue:`23539`)
 - ``max_rows`` and ``max_cols`` parameters removed from :class:`HTMLFormatter` since truncation is handled by :class:`DataFrameFormatter` (:issue:`23818`)
 - :meth:`read_csv` will now raise a ``ValueError`` if a column with missing values is declared as having dtype ``bool`` (:issue:`20591`)
+- The column order of the resultant ``DataFrame`` from ``MultiIndex.to_frame()`` is now guaranteed to match the ``MultiIndex.names`` order. (:issue:`22420`)


can you use a :meth: ref for MultiIndex.to_frame() and :attr: for .names

jreback · 2018-12-06T01:25:54Z

pandas/core/indexes/multi.py

+        3        1  jolly
+        4        2    joy
+        5        2    joy
+        >>> pd.MultiIndex.from_frame(df)


can you add a blank line between cases

jreback · 2018-12-06T01:26:29Z

pandas/core/indexes/multi.py

+                   labels=[[0, 0, 1, 1, 2, 2], [0, 1, 0, 1, 2, 2]],
+                   names=['will_be', 'used'])
+
+        >>> df = pd.DataFrame([['ahc', 'iam'], ['ahc', 'wim'], ['boh', 'amg'],


can you add a 1-line expln here (I think the first one is self-explanatorY)

jreback · 2018-12-06T01:27:18Z

pandas/tests/indexes/multi/test_constructor.py

+    np.array([[1, 2], [3, 4], [5, 6]]),
+    27
+])
+def test_from_frame_non_frame(non_frame):


rename to test_from_frame_error

jreback · 2018-12-06T01:28:09Z

pandas/tests/indexes/multi/test_conversion.py

+    # GH 22420
+    mi = pd.MultiIndex.from_arrays([
+        pd.date_range('19910905', periods=6, tz='US/Eastern'),
+        [1, 1, 1, 2, 2, 2],


is this a repeated test of the above, if so, then not necessary here.

This test was at the suggestion of @TomAugspurger
#23141 (comment)

ms7463 · 2018-12-06T04:16:13Z

@jreback - any reason why I would get this linting error in pandas-dev.pandas tests ?

Bash exited with code '1'.

Nothing else accompanying it (I fixed the other ones that were there before). The local pep8 checks don't show anything, and the branch is up to date with master.

Same thing is happening on this PR (#23538)

datapythonista · 2018-12-06T09:59:35Z

To see the error you need to click into the log. For many of them it's easy to make them red, and they are highlighted, but in this case it's an "error" with the ordering of the imports:

2018-12-06T03:44:33.5517264Z ERROR: /home/vsts/work/1/s/pandas/core/indexes/multi.py Imports are incorrectly sorted.
2018-12-06T03:44:33.5518052Z ERROR: /home/vsts/work/1/s/pandas/tests/indexes/multi/test_constructor.py Imports are incorrectly sorted.

datapythonista

Looks good, the docstring has some format errors and can be improved, but good work.

datapythonista · 2018-12-06T10:02:47Z

pandas/core/indexes/multi.py

+
+        Parameters
+        ----------
+        df : pd.DataFrame


Suggested change

df : pd.DataFrame

df : DataFrame

datapythonista · 2018-12-06T10:03:44Z

pandas/core/indexes/multi.py

+        ----------
+        df : pd.DataFrame
+            DataFrame to be converted to MultiIndex.
+        sortorder : int or None


Suggested change

sortorder : int or None

sortorder : int, optional

And please explain what it means to not be provided (None)

pandas/core/indexes/multi.py

datapythonista · 2018-12-06T10:11:08Z

pandas/core/indexes/multi.py

+                   labels=[[0, 0, 1, 1, 2, 2], [0, 1, 0, 1, 2, 2]],
+                   names=['a', 'b'])
+
+        # Use explicit names, instead of column names


Better as a paragraph than as a code comment (remove the # and leave a blank line after this line.

datapythonista · 2018-12-06T10:12:00Z

pandas/core/indexes/multi.py

+                   labels=[[0, 0, 1, 1, 2, 2], [0, 1, 0, 1, 2, 2]],
+                   names=['X', 'Y'])
+
+        See Also


This goes before the Examples section. ./scripts/validate_docstrings.py pandas.MultiIndex.from_frame should report it as an error. Make sure nothing is reported by the script.

Fixed this docstring and all the other constructor methods docstrings (since I had to modify them to update the See Alsos). I also fixed the pd.MultiIndex docstring to the best of my abilities (since I had to make some small modifications to that too). However, there were still some issues with the MultiIndex docstring:

2 Errors found: Parameters {_set_identity, name, dtype} not documented Unknown parameters {labels}

Any ideas on how I should address these? Looks like labels is there to serve as a deprecation reminder.

ms7463 · 2018-12-07T04:39:52Z

@jreback - tests are clean.

jreback · 2018-12-07T13:00:46Z

@ArtinSarraf one more merge master and looks good.

jreback · 2018-12-07T13:00:59Z

@datapythonista @toobaz if you have further comments.

datapythonista

lgtm, thanks for all the fixes to docstrings @ArtinSarraf

ms7463 · 2018-12-08T02:36:51Z

@jreback - master has been merged and tests are green.

toobaz · 2018-12-08T08:46:31Z

Looks good to me!

jreback · 2018-12-09T14:10:18Z

thanks @ArtinSarraf

sds9995 added 2 commits October 13, 2018 18:11

ENH - add from_frame method and accompanying squeeze method to multii…

79bdecb

…ndex

ENH - guarentee that order of labels is preserved in multiindex to_fr…

fa82618

…ame method

sds9995 added 2 commits October 13, 2018 18:23

CLN - adhere to PEP8 line length

64b45d6

CLN - remove trailing whitespace

64c7bb1

jreback requested changes Oct 14, 2018

View reviewed changes

sds9995 and others added 5 commits October 15, 2018 23:16

ENH - raise TypeError on inappropriate input

3ee676c

TST - add tests for mi.from_frame and mi.squeeze

fd266f5

CLN - pep8 adherence in tests

4bc8f5b

CLN - last missed pep8 fix

9d92b70

BUG - remove pd.DataFrame in favor of local import

45595ad

DOC - add more detailed docstrings for from_frame and squeeze

3530cd3

datapythonista requested changes Oct 28, 2018

View reviewed changes

DOC - update MultiIndex.from_frame and squeeze doctests to comply wit…

1c22791

…h the Pandas docstring guide

gfyoung added Enhancement MultiIndex labels Oct 28, 2018

ms7463 and others added 3 commits December 2, 2018 15:12

Merge branch 'master' into enhancement/from_frame

6960804

Merge branch 'master' into enhancement/from_frame

11c5947

Merge branch 'enhancement/from_frame' of https://github.com/ArtinSarr…

904644a

…af/pandas into enhancement/from_frame

jreback requested changes Dec 4, 2018

View reviewed changes

sds9995 added 3 commits December 4, 2018 21:20

DOC - add advanced.rst section

30fe0df

Merge branch 'master' into enhancement/from_frame

ec60563

Merge branch 'master' into enhancement/from_frame

8fc6609

jreback requested changes Dec 6, 2018

View reviewed changes

jreback added this to the 0.24.0 milestone Dec 6, 2018

sds9995 added 2 commits December 5, 2018 22:08

DOC/CLN - cleanup documentation

9b906c6

CLN - fix linting error according to pandas-dev.pandas test

e416122

datapythonista requested changes Dec 6, 2018

View reviewed changes

sds9995 added 2 commits December 6, 2018 20:58

DOC - fix docstrings

4ef9ec4

CLN - fix import order with isort

4240a1e

datapythonista approved these changes Dec 7, 2018

View reviewed changes

Merge branch 'master' into enhancement/from_frame

9159b2d

jreback approved these changes Dec 9, 2018

View reviewed changes

jreback merged commit d8fd5a6 into pandas-dev:master Dec 9, 2018

toobaz mentioned this pull request Jan 29, 2019

Un-deprecate DataFrame.from_items() #21850

Closed

Pingviinituutti pushed a commit to Pingviinituutti/pandas that referenced this pull request Feb 28, 2019

ENH: MultiIndex.from_frame (pandas-dev#23141)

c1a7d1e

Pingviinituutti pushed a commit to Pingviinituutti/pandas that referenced this pull request Feb 28, 2019

ENH: MultiIndex.from_frame (pandas-dev#23141)

fe45bc6

ENH: MultiIndex.from_frame #23141

ENH: MultiIndex.from_frame #23141

Conversation

ms7463 commented Oct 13, 2018 • edited Loading

pep8speaks commented Oct 13, 2018 • edited Loading

Comment last updated on November 19, 2018 at 04:25 Hours UTC

jreback commented Oct 13, 2018

ms7463 commented Oct 13, 2018

ms7463 commented Oct 13, 2018 • edited Loading

codecov bot commented Oct 13, 2018 • edited Loading

Codecov Report

jreback commented Oct 14, 2018

ms7463 commented Oct 14, 2018 • edited Loading

jreback left a comment

Choose a reason for hiding this comment

ms7463 commented Oct 14, 2018

jreback commented Oct 15, 2018

ms7463 commented Oct 16, 2018 • edited Loading

Setup df

Adding meta data

Quick Prototyping

Formatting MultiIndex

Complex Filtering/Slicing

Custom Sorting

ms7463 commented Oct 16, 2018

ms7463 commented Oct 20, 2018

ms7463 commented Oct 27, 2018

jreback commented Oct 28, 2018

datapythonista left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ms7463 commented Oct 28, 2018

jreback left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jreback commented Dec 4, 2018

jreback left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ms7463 commented Dec 6, 2018

datapythonista commented Dec 6, 2018

datapythonista left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ms7463 commented Dec 7, 2018 • edited Loading

jreback commented Dec 7, 2018

jreback commented Dec 7, 2018

datapythonista left a comment

Choose a reason for hiding this comment

ms7463 commented Dec 8, 2018

toobaz commented Dec 8, 2018

jreback commented Dec 9, 2018

ms7463 commented Oct 13, 2018 •

edited

Loading

pep8speaks commented Oct 13, 2018 •

edited

Loading

ms7463 commented Oct 13, 2018 •

edited

Loading

codecov bot commented Oct 13, 2018 •

edited

Loading

ms7463 commented Oct 14, 2018 •

edited

Loading

ms7463 commented Oct 16, 2018 •

edited

Loading

ms7463 commented Dec 7, 2018 •

edited

Loading