Skip to content

Latest commit

 

History

History
247 lines (169 loc) · 7.14 KB

compatibility.md

File metadata and controls

247 lines (169 loc) · 7.14 KB
Raw

#Migrating from Plotly's Python API, 0.5 to 1.0:

##Backwards Compatibility

  • import statements and module locations have changed
  • deprecated methods

careful, modules may have shifted during flight

The functionality in within plotly.plotly has been moved to two locations:

  • plotly.plotly
  • plotly.tools

Read on to see how this might affect you!

changed/moved/deprecated functionality

Some functionality has moved, or has been altogether deleted. Here's the rundown of what got effected:

Name Action
plotly.embed() moved
plotly.display() deprecated
plotly.signup() deprecated
plotly.plotly() deprecated
plotly.plotly.ion() deprecated
plotly.plotly.ioff() deprecated
plotly.plotly.plot() moved/changed
plotly.plotly.iplot() moved/changed
plotly.plotly.layout() deprecated
plotly.plotly.style() deprecated
plotly.stream.init() deprecated
plotly.stream.write() moved
plotly.stream.close() moved

###setting up credentials

Previously, you began your Plotly session by creating a plotly instance that held your credentials. You can now choose to sign_in or if you've set_credentials, you do not even need to sign in any longer.

# old way
import plotly
py = plotly.plotly('username', 'api_key')  # py is a `plotly` instance
# new way
import plotly.plotly as py  # py is now a name for the plotly.plotly module
py.sign_in('username', 'api_key')

See our guidebook for more information on logging in:

###plot and iplot call signatures:

# old way to make a plot
import plotly
py = plotly.plotly('username', 'api_key')
py.plot(data, layout=layout, **plot_options)
py.iplot(data, layout=layout, **plot_options)
# new way to make a plot
import plotly.plotly as py
py.sign_in('username', 'api_key')
py.plot(data_or_figure, **plot_options)
py.iplot(data_or_figure, **plot_options)

As before, data is a list of dictionary-like objects representing individual traces on a graph. However, layout has now been included with data in a new figure object:

# you now send a figure dictionary
figure = {'data':data, 'layout':layout}

All objects in the JSON structure you're creating can now be represented as actual Python classes.

See our guidebook for more information on the plot functions:

###'histogramx' and 'histogramy' are now just 'histogram'

This is a change in the actual JSON, but this change also effects the Python API.

# old way to make a histogram along the x-axis
option1 = dict(type='histogramx', x=[1,2,2,1,5,...,1])
option2 = dict(type='histogramy', y=[1,2,2,1,5,...,1])
# new way to make a histogram along the x-axis
trace = dict(type='histogram', x=[1,2,2,1,5,...,1])
# old way to make a histogram along the y-axis
option1 = dict(type='histogramx', x=[1,2,2,1,5,...,1], bardir='h')
option2 = dict(type='histogramy', y=[1,2,2,1,5,...,1], bardir='h')
# new way to make a histogram along the x-axis
trace = dict(type='histogram', y=[1,2,2,1,5,...,1])

'x' data is now always plotted with vertical bars and 'y' data is now always plotted with horizontal bars. When you specify 'x' and 'y', you can use 'orientation' to select whether 'x' or 'y' will be plotted in their corresponding orientations.

###specifying filename now causes fileopt='overwrite'

Previously, to overwrite a file in your plotly account, you needed to specify both 'filename' and 'fileopt'. Now, if you specify a 'filename' keyword argument when making a call to plot() or iplot(), 'fileopt' will default to 'overwrite', and any existing plot with that filename will be overwritten.

Simply specify both to set a filename, but not overwrite if that filename already exists.

###plotly.embed() has moved, and call signature has changed

# old way
import plotly
plotly.embed('https://plot.ly/~PythonAPI/67/numpy-boxes/')
# new way
import plotly.tools as tls
tls.embed('PythonAPI', 67)

Note that the location of this functionality has moved. More importantly, though, you now receive figure information from Plotly by specifying a case-insensitive string and a file_id number.

###plotly.display() is now plotly.tools.embed()

The plotly.display() function is deprecated and has been replaced with potly.tools.embed().

See the entry on embed for more details on new call signatures.

###for types 'bar' and 'histogram', 'bardir' is now 'orientation'

In general, 'orientation' now describes how x and y data relate to plots and can be used more broadly than in just 'histogram' and 'bar'. Note, however, the 'bardir' still exists in 'layout'.

###ion and ioff have been deprecated, use auto_open=True

Previously, you could set whether or not new plots created in Plotly from the API caused a new browser tab to open.

By default it will open a new tab, to stop this, you should now use the auto_open keyword:

import plotly.plotly as py
py.plot(figure, auto_open=False)

###setting layout parameters

layout is now a member of a figure dictionary. You declare layout options by setting keyword-value pairs in layout.

There are two ways to update layout options.

First, you can just add a layout dictionary into a new figure dictionary along with your data:

data = []  # put your data here
layout = {}  # set layout parameters here
figure = dict(data=data, layout=layout)  # assemble figure here

Second, if you're using the new 'graph objects' this version of the api defines, you can use the update method on a figure instance, which has been redefined to allow nested updating. You basically treat all of the new graph objects as if they were dictionaries and lists.

figure = Figure()  # auto-adds layout and data keys
figure['data'] += []  # put your data here
figure['layout'] = {}  # (option 1) set layout parameters here

# additionally, you can update in these ways
figure['layout'].update({})  # (option 2)
figure['layout'].update(key1=val1, key2=val2, key3=val3)  # (option 3)

###style() has been deprecated

All style parameters need to be declared within their resepctive objects. That is, layout styles must be declared within the layout dictionary and data object styles must be declared within their respective dictionaries.

###plotly.stream is now plotly.plotly.Stream, core functionality remains.

Aside from the class name being capitalized, the functionlity of the streaming class has improved. Run help(plotly.plotly.Stream) for more information.

###plotly.signup() has been deprecated

Users must signup initially at the plotly website, this can no longer be done through the Python API.

###Axis and axis-references better support indexing

Previously the initial xaxis key needed to be named 'xaxis'. It can now be named 'xaxis1'. Similarly references to this first axis used to required 'x', but now allow 'x1' to refer to the same axis.