Skip to contents

Migrating from plotly to myIO

Source: vignettes/plotly-migration.Rmd
plotly-migration.Rmd

Why switch from plotly?

Three things changed in 2025 that make plotly a risky dependency for new work:

  1. Documentation retired. The official plotly-r docs site went offline. Learning the API now means reading source code or archived pages.
  2. Broken statistical features. Regression CI bands (#1472) and statistical annotations (#1687) have been broken for years with no fix in sight.
  3. Heavy dependency chain. plotly pulls in htmlwidgets, tidyr, dplyr, rlang, and the full plotly.js bundle (~3.5 MB minified). myIO depends only on htmlwidgets and jsonlite.

If your charts need statistical overlays, the cost of working around plotly’s gaps exceeds the cost of switching.

Side-by-side: 8 common patterns

1. Basic scatter plot 

# plotly
plot_ly(mtcars, x = ~wt, y = ~mpg, type = "scatter", mode = "markers")
# myIO
myIO(data = mtcars) |>
  addIoLayer(type = "point", label = "Cars",
    mapping = list(x_var = "wt", y_var = "mpg"))

myIO uses a layered pipe API instead of a single function with mode flags.

2. Line chart with multiple series 

# plotly
plot_ly(economics_long, x = ~date, y = ~value, color = ~variable,
        type = "scatter", mode = "lines")
# myIO
myIO(data = economics_long) |>
  addIoLayer(type = "line", label = "Trends",
    mapping = list(x_var = "date", y_var = "value", group = "variable"))

Groups are declared in the mapping, not as a top-level aesthetic.

3. Bar chart 

# plotly
plot_ly(data.frame(x = c("A","B","C"), y = c(10,20,15)),
        x = ~x, y = ~y, type = "bar")
# myIO
myIO(data = data.frame(x = c("A","B","C"), y = c(10,20,15))) |>
  addIoLayer(type = "bar", label = "Values",
    mapping = list(x_var = "x", y_var = "y")) |>
  defineCategoricalAxis(xAxis = TRUE)

myIO requires an explicit defineCategoricalAxis() call for discrete x-axes.

4. Histogram 

# plotly
plot_ly(mtcars, x = ~mpg, type = "histogram")
# myIO
myIO(data = mtcars) |>
  addIoLayer(type = "histogram", label = "MPG Distribution",
    mapping = list(x_var = "mpg"),
    options = list(bins = 15))

Bin count is set via options$bins rather than a layout parameter.

5. Box plot 

# plotly
plot_ly(iris, y = ~Sepal.Length, color = ~Species, type = "box")
# myIO
myIO(data = iris) |>
  addIoLayer(type = "boxplot", label = "Sepal Length",
    mapping = list(x_var = "Species", y_var = "Sepal.Length"),
    options = list(showOutliers = TRUE)) |>
  defineCategoricalAxis(xAxis = TRUE)

myIO boxplots decompose into sub-layers (IQR box, whiskers, median, outliers), each independently styled and interactive.

6. Regression line with CI band

plotly #1472 – CI ribbons on regression lines do not render correctly.

# plotly (broken — CI band misaligns or disappears)
model <- lm(mpg ~ wt, data = mtcars)
preds <- data.frame(wt = seq(min(mtcars$wt), max(mtcars$wt), length.out = 50))
preds <- cbind(preds, predict(model, preds, interval = "confidence"))
plot_ly() |>
  add_markers(data = mtcars, x = ~wt, y = ~mpg) |>
  add_ribbons(data = preds, x = ~wt, ymin = ~lwr, ymax = ~upr) |>
  add_lines(data = preds, x = ~wt, y = ~fit)
# myIO (one call, CI computed internally)
myIO(data = mtcars) |>
  addIoLayer(type = "regression", label = "MPG vs Weight",
    mapping = list(x_var = "wt", y_var = "mpg"),
    options = list(method = "lm", showCI = TRUE, showStats = TRUE))

myIO computes the CI via stats::predict() and renders it as a first-class area layer – no manual pre-computation.

7. Statistical annotations

plotly #1687 – ggplotly() drops stat_compare_means() annotations.

# plotly (annotations lost in ggplotly conversion)
library(ggpubr)
p <- ggboxplot(iris, x = "Species", y = "Sepal.Length") +
  stat_compare_means(method = "t.test", comparisons = list(
    c("setosa", "versicolor"), c("versicolor", "virginica")))
ggplotly(p)  # brackets and p-values vanish
# myIO (pairwise tests rendered natively)
myIO(data = iris) |>
  addIoLayer(type = "comparison", label = "Sepal Length",
    mapping = list(x_var = "Species", y_var = "Sepal.Length"),
    options = list(method = "t.test"))

The comparison composite expands into boxplots plus significance brackets with p-values, computed in R and rendered in D3.js.

8. Dark mode theming 

# plotly
plot_ly(mtcars, x = ~wt, y = ~mpg, type = "scatter", mode = "markers") |>
  layout(template = "plotly_dark")
# myIO
myIO(data = mtcars) |>
  addIoLayer(type = "point", label = "Cars",
    mapping = list(x_var = "wt", y_var = "mpg")) |>
  setTheme(background = "#1a1a2e", text = "#e0e0e0",
           grid = "#2a2a4a", font = "Inter")

myIO theming uses CSS custom properties, so colors apply consistently across all layers including CI bands, annotations, and export buttons.

What myIO does that plotly can’t

  • Composite charts. regression, comparison, qq, violin, and ridgeline auto-expand into coordinated sub-layers.
  • Composable transforms. lm, loess, ci, mean_ci, residuals, pairwise_test, and qq mix freely across layers.
  • Bidirectional I/O. setBrush() returns selected rows; setAnnotation() enables click-to-label with CSV export.

What plotly does that myIO doesn’t

  • 3D charts. scatter3d, surface, mesh3d – myIO is 2D only.
  • Geographic maps. scattergeo, choropleth, Mapbox – myIO has no map types.
  • ggplotly conversion. No equivalent one-liner for converting existing ggplot2 code.
  • 40+ chart types. Funnel, icicle, sunburst, parallel coordinates, etc. myIO has 23 types focused on statistical workflows.

If you need 3D, maps, or broad chart-type coverage, plotly remains the better choice. If you need statistical overlays that actually work, myIO is worth the switch.