Using ginput on a plot_date plot doesn't return datetime

When using ginput() on a plot made using plot_date(), the return type of the x value is float, but it should be datetime.

Minimal example:

import matplotlib.pyplot as plt
from datetime import datetime, timedelta

times = []
data = []
for i in range(0, 10):
    times.append(datetime(2016, 1, 1, 0, 0, 0) + timedelta(seconds=i))
    data.append(i)

plt.plot_date(times, data)
x = plt.ginput()
print(x[0][0])
print(type(x[0][0]))

The last statement prints "<class 'numpy.float64'>"

Present on 2.0.0b4+2828.g86b32ca (installed from source using pip on Linux)

Not sure if we actually know enough to be able to do this in general right now.

datetime is handled through the unit framework which (roughly) works by:

registering a function that converts units -> float
registering a formatter that converts float -> string representation of value

I think you are asking to also provide an inverse from float -> the input data type with units.

It should be possible to add this in a back compatible way, but will need a bit of thought.

For this particular case, iirc the float is ~~seconds from 1900~~ days from 0001-01-01 UTC, plus 1.

It's days from a ridiculous but common sort of origin, long, long, ago. See http://matplotlib.org/api/dates_api.html for the explanation.

Would it not be possible to use num2date to convert the output of ginput directly to a datetime? http://matplotlib.org/api/dates_api.html#matplotlib.dates.num2date

@tacaswell

@tacaswell, can you please provide more information on the unit framework? Where in the code is it? I traced the function calls to BlockingMouseInput but I don't understand exactly how and where the mouse click coordinates are retrieved. Perhaps some information on that might also help me.

added unit tests file for issue matplotlib#7462

@tacaswell

After thorough investigation of this bug as a team for our project at UofT, we came to the conclusion there isn't a really good solution to this problem.

We had the following fix in mind:

introduce a new parameter to ginput called converter which we will use to pass the desired converter function to ginput.
the converter function passed must take a float
The output of ginput() will now be the output of the converter function or a float if no converter is passed in.

Advantages of the fix:
1.The float that ginput() generates, can be converted to any type as long as the user provides a converter function.
2. You can technically pass in any function in place of the converter and use it to deduce information about the data

Disadvantages of the fix:

The output type of ginput() is inconsistent. This is a sign of bad software design although python permits it.
No real change to what's available now. The only thing is that the conversion step is done implicitly. This however, introduces the inconsistency in the output type of ginput().

Conclusion
The current design of ginput() is the most appropriate one. Conversions should be done explicitly to preserve the consistency of the output type of ginput(). Perhaps there may be another way to work this out by introducing polymorphism for all the possible types that could be used in a plot and change ginput to have an output of the parent type, however that will require quite a bit of restructuring and the benefit may not be worth the effort.

@tacaswell, do you still consider this a feature that's worth implementing? If not, can we close this issue? Thanks

I think the best approach would be to add an example that uses ginput + num2date to get out the desired datatime with some comments about how if you are using unitful data + ginput the user is responsible for taking care of the inverse transformation.

@tacaswell

Thanks @tacaswell. Under what subsection of examples should we add this example? Also, are we responsible for adding the HTML pages and links for the new example we will create?

This issue has been marked "inactive" because it has been 365 days since the last comment. If this issue is still present in recent Matplotlib releases, or the feature request is still wanted, please leave a comment and this label will be removed. If there are no updates in another 30 days, this issue will be automatically closed, but you are free to re-open or create a new issue if needed. We value issue reports, and this procedure is meant to help us resurface and prioritize issues that have not been addressed yet, not make them disappear. Thanks for your help!

I think most converters have inverses? If so we could just have an option to ginout to change the output type. The inverse isn't unique, but could at least be in the units/type the user wants. Note that the cursor status in the lower right already does this.

Marking as good first issue. Basically the following should be done:

Add a parameter to ginput that makes a conversion to the input format (through the converter inverse) with documentation

Figure out how to get that value, looking at e.g. the status message formatting:

matplotlib/lib/matplotlib/backend_bases.py

Lines 2941 to 2960 in 8ca75e4

    
           def _mouse_event_to_message(event): 
        
               if event.inaxes and event.inaxes.get_navigate(): 
        
                   try: 
        
                       s = event.inaxes.format_coord(event.xdata, event.ydata) 
        
                   except (ValueError, OverflowError): 
        
                       pass 
        
                   else: 
        
                       s = s.rstrip() 
        
                       artists = [a for a in event.inaxes._mouseover_set 
        
                                  if a.contains(event)[0] and a.get_visible()] 
        
                       if artists: 
        
                           a = cbook._topmost_artist(artists) 
        
                           if a is not event.inaxes.patch: 
        
                               data = a.get_cursor_data(event) 
        
                               if data is not None: 
        
                                   data_str = a.format_cursor_data(data).rstrip() 
        
                                   if data_str: 
        
                                       s = s + '\n' + data_str 
        
                       return s 
        
               return ""

Add a test

Note that one will have to dig a bit deeper, as we are not really interested in the string formatting of the cursor position here.

Hi! I wanted to work on this. I have a question -- since ginput returns (x, y), should we pass two separate converters in case the axes are using different units (for example plotting datetime against floats, etc)?

@tacaswell

Looking back through the comments, the recommendation from @tacaswell was that this option should not be added to ginput. Instead, an example should be added to the documentation showing how to convert the output of ginput back to the desired units. That was 5 years ago. To be clear: is the reversal of this guidance coming because of changes in mpl in the interim, or is it just a different opinion?

Navigation Menu

Repository navigation

Using ginput on a plot_date plot doesn't return datetime #7462

7 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

	def _mouse_event_to_message(event):
	if event.inaxes and event.inaxes.get_navigate():
	try:
	s = event.inaxes.format_coord(event.xdata, event.ydata)
	except (ValueError, OverflowError):
	pass
	else:
	s = s.rstrip()
	artists = [a for a in event.inaxes._mouseover_set
	if a.contains(event)[0] and a.get_visible()]
	if artists:
	a = cbook._topmost_artist(artists)
	if a is not event.inaxes.patch:
	data = a.get_cursor_data(event)
	if data is not None:
	data_str = a.format_cursor_data(data).rstrip()
	if data_str:
	s = s + '\n' + data_str
	return s
	return ""

Using ginput on a plot_date plot doesn't return datetime #7462

Description

Activity

tacaswell commented on Nov 20, 2016

efiring commented on Nov 20, 2016

dstansby commented on Feb 28, 2017

mitko1994 commented on Mar 16, 2017

mitko1994 commented on Apr 3, 2017

tacaswell commented on Apr 3, 2017

mitko1994 commented on Apr 10, 2017

7 remaining items

github-actions commented on Apr 5, 2023

jklymak commented on Apr 5, 2023

oscargus commented on Apr 5, 2023

turnipseason commented on May 29, 2023

efiring commented on May 29, 2023

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions