Fixed missing marker in legend when plotting a series whose first entry is NaN

[leespen1]

Changed function gr_add_legend to call gr_draw_marker instead of gr_draw_markers, since a legend entry only has one marker (if any).

Fix #3996

I explained the issue and solution in greater detail in this post: https://discourse.julialang.org/t/fixing-missing-marker-in-plot-legend-when-first-value-in-series-is-nan/86658

In summary, the example

using Plots; gr()
plot([NaN,2,3], markershape=:circle)

produced a graph with no marker in the legend:

This only happens when the first entry in the series is NaN. If only the second entry is NaN, we get:

using Plots; gr()
plot([1,NaN,3], markershape=:circle)

Morever, this behavior does not happen for the PlotlyJS backend (I did not test backends besides GR and PlotlyJS):
he second entry is NaN, we get:

using Plots; plotlyjs()
plot([NaN,2,3], markershape=:circle)

By changing the gr_draw_markers call to a gr_draw_marker call, this behavior was fixed. Now the example

using Plots; gr()
plot([NaN,2,3], markershape=:circle)

produces the following graph:

And it works for a single series with multiple marker shapes as well:

using Plots; gr()
plot([1,2,3], markershape=[:circle, :square, :diamond])

The only downside I can see is the following example:

using Plots; gr()
plot([NaN,2,3], markershape=[:circle, :square, :diamond])

The marker in the legend is a circle, although no circle appears in the series because the first entry in the series is not plotted (because it's NaN). Still, the previous behavior was to have no marker at all, which I think is worse.

[t-bltg]

You might want to read https://docs.juliaplots.org/stable/contributing/#Write-code,-and-format.

Also, please rebase your PR against current master for CI to run without errors.

[codecov]

Codecov Report

Base: 80.23% // Head: 81.05% // Increases project coverage by +0.81% 🎉

Coverage data is based on head (c6af026) compared to base (3f8904a).
Patch coverage: 89.43% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #4328      +/-   ##
==========================================
+ Coverage   80.23%   81.05%   +0.81%     
==========================================
  Files          29       28       -1     
  Lines        7009     7316     +307     
==========================================
+ Hits         5624     5930     +306     
- Misses       1385     1386       +1     

Impacted Files	Coverage Δ
src/Plots.jl	87.80% <ø> (ø)
src/recipes.jl	67.42% <0.00%> (+1.35%)	⬆️
src/output.jl	62.38% <33.33%> (-0.58%)	⬇️
src/utils.jl	75.62% <66.66%> (+0.29%)	⬆️
src/layouts.jl	79.77% <75.00%> (+3.07%)	⬆️
src/types.jl	88.67% <85.00%> (+6.86%)	⬆️
src/backends/unicodeplots.jl	87.55% <90.90%> (+2.07%)	⬆️
src/backends/gr.jl	89.02% <96.42%> (+0.44%)	⬆️
src/args.jl	80.39% <100.00%> (+0.72%)	⬆️
src/axes.jl	85.83% <100.00%> (-0.60%)	⬇️
... and 28 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[leespen1]

Every check passed except "ci / Julia 1.6 - macos-latest (pull_request)". Does this mean the fix (or the changes made to the original project) is not compatible with Julia1.6 on MacOS? I use Ubuntu, so I'm not sure what I should do about this.

[t-bltg]

This is just a spurious random error related to network failure : GitError(Code:ERROR, Class:Net, SecureTransport error: connection closed via error).

I'm re-running the failed job now, don't worry this is not blocking.

[t-bltg]

I think that this PR is well explained (here and on discourse), and I see no regressions in the examples on ci, so I don't see any reason why this shouldn't go in.

[BeastyBlacksmith]

Thank you for this fix! You might want to check out #3503.

[leespen1]

You're welcome, thanks for the review!