Use light background in plots?

[cpascual]

Intro

Pyqtgraph's choice of a dark background for plots has generated mixed reactions among taurus users.

Changing it to a light background has been a common request, but the decision has always been postponed because of lack of time and its not-so-obvious implications.

But it may be good to do some discussion about it now in order to take decision.

Note: pyqtgraph plot widgets allow calling .setBackgroundBrush() individually, but pyqtgraph as a whole provides a global switch mechanism.

Arguments for each option

Apart of purely aesthetic preferences (there is people who prefers dark themes and others who prefer light ones), here are some arguments for each option:

Arguments in favor of changing background to a light color:

• It makes the curves easier to see (but whether this is intrinsic or it can be managed with a change in the colors for the curves palette, is not clear to me)
• It is more similar to the qwt5-based plots that we are replacing (less change stress on the users, smoother transition)
• It matches better the default desktop theme in KDE (which is light)

Arguments in favor of keeping the original dark background :

• It is less work: changing the colors is easy, but providing configurability that covers all use cases may be difficult (see next section)
• Concerns that "going against upstream" will result in more maintenance work and possibly require more tweaks by end users (e.g. it is possible that some design decisions assume that colors are the default dark ones, and we or the final user will need to do extra work to fix issues that would not be apparent if we used the default)
• It maintains visual consistency with pure-pyqtgraph widgets. Note that a design goal of taurus_pyqtgraph is to encourage the use of pure pyqtgraph widgets and APIs and minimize the amount of taurus-specific knowledge needed to use it. Nevertheless, if we use the global switch mechanism for changing the colors at pyqtgraph level, all pyqtgraph-based widgets in the same application would be affected

Configurability

Whichever the decision, should we provide a simple "switch" mechanism?

And if so, which scope should this mechanism act onto (a global configuration, at the widget / tool level, ...)?

And should the mechanism be used implicitly (e.g. set the colors when importing taurus_pyqtgraph) or explicitly (e.g. a function that you need to call in all your guis).

Note that an implicit and global change may seem convenient but will probably upset GUIs which mix pure pyqtgraph widgets with taurus_pyqtgraph ones.

[cpascual]

Here is a comparison of different options:

• upper left: black background (current for tpg plot, which uses the default in pyqtgraph)
• upper right: old qwt5 plot (for reference)
• bottom left: white background
• bottom right: light gray background (mimics qwt5)

[cpascual]

In the above screenshot, it is evident that the current choice of the curve colors is more appropriate for a light background than a dark one. Here is another screenshot where I manually changed curve colors for better contrast (changing the default palette of curve colors is something we definitely should do if we keep the dark background).

[cpascual]

Note that since v0.4.0, the background color for each plot can be set via plot config dialog and it is (re)stored via config properties.

[reszelaz]

Great news! IMO it is the best possible option to let the user to choose. Many thanks!
Have you left the configuration per trend/plot or also a global one?

[cpascual]

Have you left the configuration per trend/plot or also a global one?

No, this is a related, but independent setting at widget level. IMHO, using a global switch (and adapting the curve color palette accordingly) is still something we should consider