Date fixes #3947

pganssle · 2014-12-25T00:09:00Z

Here are some fixes to some of the problems discussed in Issue #2845. Sorry I worked off of the master branch.

I consolidated all the "magic numbers" into a central location so they are consistent and easy to update if necessary, and I updated the to_ordinalf and from_ordinalf functions to use datetime.timedelta rather than deal with converting fractional days to datetime objects and vice versa.

…triple-quote docstrings for consistency.) Signed-off-by: Paul G <p.ganssle@gmail.com>

Signed-off-by: Paul G <p.ganssle@gmail.com>

…or the moment. Signed-off-by: Paul G <p.ganssle@gmail.com>

…om `datetime`. Signed-off-by: Paul G <p.ganssle@gmail.com>

…nversion. Signed-off-by: Paul G <p.ganssle@gmail.com>

pganssle · 2014-12-25T00:15:47Z

One thing to note is that in the current master branch, dates.py is inconsistent between how long a "month" is. In some places it's 31 days, in other places 30 days. By switching all these references to DAYS_PER_MONTH, I set them all to 30. It seems minor, but that seems like the main thing that might change behavior.

I'll have to look through what the results would be, but I'm wondering if it might be a good idea to set DAYS_PER_MONTH to be DAYS_PER_YEAR/MONTHS_PER_YEAR, which would be 365.0 / 12.0 or 30.41666. I'm not sure what the consequences of that would be. Similarly, since I've changed all references to the number of days per year from 365.0 to DAYS_PER_YEAR, I'm wondering if it would be more accurate to set DAYS_PER_YEAR to some number closer to an actual year, like 365.242.

Signed-off-by: Paul G <p.ganssle@gmail.com>

…atplotlib#3896. Signed-off-by: Paul G <p.ganssle@gmail.com>

…ects. Signed-off-by: Paul G <p.ganssle@gmail.com>

Signed-off-by: Paul G <p.ganssle@gmail.com>

pganssle · 2014-12-26T04:36:42Z

Sorry these builds keep failing - I don't have my full normal set up at the moment and its a bit of a pain to set up the Windows builds, so I can't do much more than basic unit tests. I also don't have my normal linting set up. I'll try and get everything sorted soon and make sure it passes the build tests.

tacaswell · 2014-12-26T13:38:56Z

Thanks for working on this.

I think master branch is the right place for these changes.

It looks like at least part of the problem is that the converted times are now coming out with a timezone attached to them.

Signed-off-by: Paul G <p.ganssle@gmail.com>

jbmohler · 2014-12-26T22:42:27Z

@pganssle It shouldn't be a pain to build on windows using https://github.com/jbmohler/matplotlib-winbuild . I'm paying attention to any gotcha's on that.

pganssle · 2014-12-26T22:50:37Z

@jbmohler Thanks for the info. I saw that, but I don't have Visual Studio installed on this computer and it seems I have to register with Microsoft to get it, which I'm not really interested in doing. I'll either bite the bullet and register or I'll set up a virtual linux container here, because at the moment I've got something rigged up with cygwin that seems to just be limping along.

…till fail, but that's fine for now. Signed-off-by: Paul G <p.ganssle@gmail.com>

jbmohler · 2014-12-27T00:07:20Z

@pganssle I do not believe that you need to register for the express edition of 2008 (Python 2.7) or 2010 (Python 3.3/4). I think you cannot build 64 bit with the express edition.

…ys between two points. Signed-off-by: Paul G <p.ganssle@gmail.com>

pganssle · 2014-12-27T00:58:48Z

OK, so I've cleared up most of the problems with the build failing. One that still remains arises from the fact that I've actually changed the behavior of AutoDateLocator.get_locator(), which was actually potentially incorrectly calculating the number of days between two events, because it needed to make an estimate of the number of days in a year and the number of months in a year. This will cause tests.test_dates.test_auto_locator() to fail, since that asserts that the date location logic hasn't changed (which it has, slightly). I'm thinking that the solution to this is to just change the values of the assert, since the expected behavior is changing, but I didn't want to presume to do that myself.

~~I'm not 100% clear on why tests.test_dates.test_DateFormatter() is failing.~~

~~This is the expected result:~~

~~And this is the actual result:~~

~~I'm guessing something I've done changes the direction or sign of some rounding errors, which cascades into this, but it doesn't seem like a _problem_, per se. I'll try and figure out exactly what's causing it so we can decide whether this new behavior is the result of something that might have other consequences. My guess is there won't be any significant consequences, but maybe while tracking it down I can figure out why the auto locator isn't choosing round numbers for fractional seconds in the first place (i.e. 0.2, 0.4, 0.6, 0.8), as that seems like it would be a preferable behavior anyway.~~ **Edit** Nevermind - seems to just be a difference between my build environment and the standard one, since that test is no longer failing on the Travis CI builds.

…lta object. Signed-off-by: Paul G <p.ganssle@gmail.com>

…timedelta` object actually calculates those for us as is. Signed-off-by: Paul G <p.ganssle@gmail.com>

Signed-off-by: Paul G <p.ganssle@gmail.com>

…e date locator. Signed-off-by: Paul G <p.ganssle@gmail.com>

…vious versions. Signed-off-by: Paul G <pg@example.com>

Signed-off-by: Paul G <pg@example.com>

pganssle · 2014-12-29T07:38:29Z

@tacaswell Good call. Was thinking of it as an instance method, but it makes more sense to think of it as an alias to the static call anyway.

…hen dateutil <= 2.3 Signed-off-by: Paul G <pg@example.com>

…en necessary. Signed-off-by: Paul G <pg@example.com>

pganssle · 2015-01-30T20:17:37Z

By the way, I know I mentioned on the dev list that I was going to look into using numpy's datetime64, but I haven't had time yet, and these date fixes are pretty much ready to go independent of that.

Please let me know if it needs to be rebased or if you want me to squash some of these commits or whatever, or anything else I need to fix, I suppose.

tacaswell · 2015-01-30T20:21:35Z

Thanks for the ping, this had fallen off my radar.

@mdboom How is the expert dates?

tacaswell · 2015-01-30T20:23:31Z

lib/matplotlib/tests/test_dates.py

-                 '1990-02-17 00:00:00+00:00', '1990-03-10 00:00:00+00:00',
-                 '1990-03-31 00:00:00+00:00', '1990-04-21 00:00:00+00:00',
-                 '1990-05-12 00:00:00+00:00']
+               [datetime.timedelta(days=141),


I remember talking about this before, but not the conclusion. Why is it OK that the tests were changed here?

@tacaswell It's an expected (minor) behavior change caused by an increase in the accuracy of the way marker locations are calculated. I think the change was in this commit; previously if the beginning and end dates were January 3rd, 2000 and June 10th, 2015, relativedelta object returns (years=15, months=5, days=7), and when get_locator() is called, it calculated the number of days between the dates assuming that the number of days per year is 365 and the number of days per month is 30, so the number of days calculated would be 5632 rather than the true value of 5637.

It's a small inaccuracy, but it's easy to fix by holding on to a timedelta object and using that when calculating days from years or months. As a result, some of the date range calculations are slightly different (though not to a degree that anyone would really notice), so at least one of the tests had to be updated (I vaguely remember updating one that was around 3 months and one that was around 3 weeks).

(Oops, responded to this in the PR not the line)

tacaswell · 2015-06-16T03:24:53Z

lib/matplotlib/dates.py

+
+        Currently `datetime.datetime.strftime()` is not supported for
+        years <= 1900 in Python 2.x and years <= 1000 in Python 3.x. This
+        function extends this functionality to


This cuts off rather abruptly

Good catch. I'll fix this a bit later this week.

Does not matter, removed in favor of #3242's changes

tacaswell · 2015-06-16T03:27:29Z

@pganssle Two super minor comments. This look almost ready!

WeatherGod · 2015-07-12T13:35:23Z

@pganssle, this needs a rebase, and address @tacaswell 's two minor comments.

tacaswell · 2015-07-16T04:18:32Z

This also needs a rebase.

pganssle · 2015-07-16T10:46:23Z

Can do. Just started a new job, so I had to sort moving and IP issues. No
problems, so I should be able to get to it this weekend.
On Jul 16, 2015 12:18 AM, "Thomas A Caswell" notifications@github.com
wrote:

This also needs a rebase.

—
Reply to this email directly or view it on GitHub
#3947 (comment)
.

tacaswell · 2015-07-16T10:59:34Z

I know how that goes, congratulations!

On Thu, Jul 16, 2015, 6:46 AM Paul Ganssle notifications@github.com wrote:

Can do. Just started a new job, so I had to sort moving and IP issues. No
problems, so I should be able to get to it this weekend.
On Jul 16, 2015 12:18 AM, "Thomas A Caswell" notifications@github.com
wrote:

This also needs a rebase.

—
Reply to this email directly or view it on GitHub
<
https://github.com/matplotlib/matplotlib/pull/3947#issuecomment-121825037>
.

—
Reply to this email directly or view it on GitHub
#3947 (comment)
.

tacaswell · 2015-07-22T03:22:27Z

This has major conflicts with the changes merged in #3242

pganssle · 2015-07-22T03:35:35Z

Yes, I noticed this when trying to rebase this past weekend. Working on resolving the conflicts.

Merged locally to deal with conflicts arising from #3242 (merged as 8609886) which also touched the pre-1900 date formatting code. Resolved conflicts in favor of #3242 which refactored the pre-1900 code into a helper function.

tacaswell · 2015-07-22T03:37:14Z

@pganssle Ah, sorry didn't realize you were working on this! I just did the merge/conflict resolution and all the tests pass locally....

pganssle · 2015-07-22T03:38:35Z

@tacaswell Oh, no problem if it works. I had done the rebase but couldn't get the tests to run on my laptop (Windows), so I was just trying to transfer it over to a linux virtual box to test it. If the tests pass then I'm fine with that.

pganssle added 5 commits December 24, 2014 09:24

Clean up docstrings (convert several single-quote string comments to …

621a840

…triple-quote docstrings for consistency.) Signed-off-by: Paul G <p.ganssle@gmail.com>

Small expansions to docstrings in the date library.

504044e

Signed-off-by: Paul G <p.ganssle@gmail.com>

Moved all the "magic numbers" into some centrally defined constants f…

390c1f4

…or the moment. Signed-off-by: Paul G <p.ganssle@gmail.com>

Replace hard-coded conversion logic in _to_ordinalf with functions fr…

5af753c

…om `datetime`. Signed-off-by: Paul G <p.ganssle@gmail.com>

Make from_ordinalf use mostly functions from datetime for time co…

ee7931c

…nversion. Signed-off-by: Paul G <p.ganssle@gmail.com>

tacaswell added the status: needs review label Dec 25, 2014

pganssle mentioned this pull request Dec 25, 2014

Why are we doing second -> date conversion our selves? #2845

Closed

pganssle added 4 commits December 24, 2014 18:38

Fix issue causing failed builds.

4b0d94b

Signed-off-by: Paul G <p.ganssle@gmail.com>

Fix problem with inconsistent time zones. This now also fixes Issue m…

ba623a7

…atplotlib#3896. Signed-off-by: Paul G <p.ganssle@gmail.com>

Didn't realize that _to_ordinalf needs to support datetime.date obj…

9f66ae2

…ects. Signed-off-by: Paul G <p.ganssle@gmail.com>

PEP8 fix.

d61397d

Signed-off-by: Paul G <p.ganssle@gmail.com>

pganssle added 2 commits December 26, 2014 15:07

Fix timezone issue, update code to coding standards.

0a5fbc6

Signed-off-by: Paul G <p.ganssle@gmail.com>

Fix timezone problem, clarify docstring for timezones.

72ef748

Signed-off-by: Paul G <p.ganssle@gmail.com>

Fixed issue with _to_ordinalf(), should work fine now. Build will s…

7fd2d36

…till fail, but that's fine for now. Signed-off-by: Paul G <p.ganssle@gmail.com>

Change behavior of get_locator to accurately calculate number of da…

2f80916

…ys between two points. Signed-off-by: Paul G <p.ganssle@gmail.com>

Absolute value must apply to timedelta object as well as reslative de…

ca284d1

…lta object. Signed-off-by: Paul G <p.ganssle@gmail.com>

pganssle force-pushed the date_fixes branch from dc7351c to ca284d1 Compare December 27, 2014 01:46

pganssle and others added 6 commits December 26, 2014 19:58

Adjust calculation of numSeconds and numMicroseconds, since the `…

68f3018

…timedelta` object actually calculates those for us as is. Signed-off-by: Paul G <p.ganssle@gmail.com>

Another adjustment to allow external libraries to do the calcs for us.

88e6e1e

Signed-off-by: Paul G <p.ganssle@gmail.com>

Adjusted test_dates to account for the minor change in behavior of th…

0de1e84

…e date locator. Signed-off-by: Paul G <p.ganssle@gmail.com>

Adjusted rounding logic in _from_ordinalf to be consistent with pre…

f585a17

…vious versions. Signed-off-by: Paul G <pg@example.com>

Add alias for total_seconds() to support Python 2.6

2a0865b

Signed-off-by: Paul G <pg@example.com>

Cleaner way to test for existence of total_seconds().

f1f6d8f

Signed-off-by: Paul G <pg@example.com>

pganssle mentioned this pull request Dec 30, 2014

bug: inconsistent types accepted in DateLocator subclasses #3897

Closed

Support numpy arrays in DayLocator, WeekdayLocator and MonthLocator w…

35b93d2

…hen dateutil <= 2.3 Signed-off-by: Paul G <pg@example.com>

pganssle force-pushed the date_fixes branch from e09c111 to 35b93d2 Compare December 31, 2014 00:14

Modified strftime code so it only falls back to this other routine wh…

ad55256

…en necessary. Signed-off-by: Paul G <pg@example.com>

tacaswell added this to the v1.5.x milestone Jan 30, 2015

tacaswell reviewed Jan 30, 2015
View reviewed changes

tacaswell mentioned this pull request Feb 7, 2015

dates.date2num bug with daylight switching hour #3896

Closed

tacaswell modified the milestones: 1.5.0, v1.5.x Feb 7, 2015

tacaswell reviewed Jun 16, 2015
View reviewed changes

tacaswell added status: needs revision and removed status: needs review labels Jul 16, 2015

tacaswell modified the milestones: proposed next point release, next point release Jul 22, 2015

tacaswell merged commit ad55256 into matplotlib:master Jul 22, 2015

tacaswell removed the status: needs revision label Jul 22, 2015

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Date fixes #3947

Date fixes #3947

pganssle commented Dec 25, 2014

pganssle commented Dec 25, 2014

pganssle commented Dec 26, 2014

tacaswell commented Dec 26, 2014

jbmohler commented Dec 26, 2014

pganssle commented Dec 26, 2014

jbmohler commented Dec 27, 2014

pganssle commented Dec 27, 2014

pganssle commented Dec 29, 2014

pganssle commented Jan 30, 2015

tacaswell commented Jan 30, 2015

tacaswell Jan 30, 2015

pganssle Jan 30, 2015

tacaswell Jun 16, 2015

pganssle Jun 16, 2015

tacaswell Jul 22, 2015

tacaswell commented Jun 16, 2015

WeatherGod commented Jul 12, 2015

tacaswell commented Jul 16, 2015

pganssle commented Jul 16, 2015

tacaswell commented Jul 16, 2015

tacaswell commented Jul 22, 2015

pganssle commented Jul 22, 2015

tacaswell commented Jul 22, 2015

pganssle commented Jul 22, 2015

Date fixes #3947

Date fixes #3947

Conversation

pganssle commented Dec 25, 2014

pganssle commented Dec 25, 2014

pganssle commented Dec 26, 2014

tacaswell commented Dec 26, 2014

jbmohler commented Dec 26, 2014

pganssle commented Dec 26, 2014

jbmohler commented Dec 27, 2014

pganssle commented Dec 27, 2014

pganssle commented Dec 29, 2014

pganssle commented Jan 30, 2015

tacaswell commented Jan 30, 2015

tacaswell Jan 30, 2015

Choose a reason for hiding this comment

pganssle Jan 30, 2015

Choose a reason for hiding this comment

tacaswell Jun 16, 2015

Choose a reason for hiding this comment

pganssle Jun 16, 2015

Choose a reason for hiding this comment

tacaswell Jul 22, 2015

Choose a reason for hiding this comment

tacaswell commented Jun 16, 2015

WeatherGod commented Jul 12, 2015

tacaswell commented Jul 16, 2015

pganssle commented Jul 16, 2015

tacaswell commented Jul 16, 2015

tacaswell commented Jul 22, 2015

pganssle commented Jul 22, 2015

tacaswell commented Jul 22, 2015

pganssle commented Jul 22, 2015