Can we use gghighlight() for autoplot()?

[hoxo-m]

Hi, thank you for the greatest package.
Now, I'm trying to use gghighlight() for a result of autoplot(), but it occurs an error.

For instance:

library(glmnetUtils)
library(ggfortify)

model <- glmnet(medv ~ ., data = MASS::Boston)

g <- autoplot(model, xvar = "lambda")
g

library(gghighlight)

g + gghighlight(variable == "crim", use_group_by = FALSE)

Error in mutate_impl(.data, dots) :
Evaluation error: object 'variable' not found

head(g$data)
#>  Df   Lambda  dev.ratio    index variable value
#> 1  0 6.777654 0.00000000 1.913631     crim     0
#> 2  1 6.175546 0.09238648 1.820597     crim     0
#> 3  2 5.626927 0.17376616 1.727563     crim     0
#> 4  2 5.127046 0.25269182 1.634530     crim     0
#> 5  2 4.671574 0.31820575 1.541496     crim     0
#> 6  2 4.256564 0.37259654 1.448462     crim     0

What would you suggest I do?

[yutannihilation]

Thanks, confirmed.

The reason of the error is that the data of the second layer doesn't contain variable.

library(glmnetUtils)
library(ggfortify)
#> Loading required package: ggplot2

model <- glmnet(medv ~ ., data = MASS::Boston)

g <- autoplot(model, xvar = "lambda")

lapply(g$layers, function(l) {
  if (is.data.frame(l$data)) {
    head(l$data)
  } else {
    l$data
  }
})
#> [[1]]
#> list()
#> attr(,"class")
#> [1] "waiver"
#> 
#> [[2]]
#>     crim zn indus     chas nox        rm age dis rad tax    ptratio
#> s21    0  0     0 0.128706   0 3.8961005   0   0   0   0 -0.6305235
#> s17    0  0     0 0.000000   0 3.5480395   0   0   0   0 -0.5076149
#> s14    0  0     0 0.000000   0 3.2365996   0   0   0   0 -0.3713991
#> s10    0  0     0 0.000000   0 2.6603223   0   0   0   0 -0.1192839
#> s7     0  0     0 0.000000   0 1.9756431   0   0   0   0  0.0000000
#> s3     0  0     0 0.000000   0 0.5694424   0   0   0   0  0.0000000
#>           black      lstat Df    Lambda dev.ratio       index  label_y
#> s21 0.002301817 -0.4975636  5 0.9607149 0.6653606 -0.04007759 4.272373
#> s17 0.000000000 -0.4828300  3 1.3938328 0.6418216  0.33205737 4.272373
#> s14 0.000000000 -0.4541851  3 1.8425650 0.6143107  0.61115860 4.272373
#> s10 0.000000000 -0.4011381  3 2.6732463 0.5432586  0.98329356 4.272373
#> s7  0.000000000 -0.3353772  2 3.5338742 0.4552422  1.26239478 4.272373
#> s3  0.000000000 -0.1969814  2 5.1270464 0.2526918  1.63452975 4.272373

^{Created on 2018-09-21 by the reprex package (v0.2.1)}

Unfortunately, currently there's no way to apply gghighlight() only on the first layer due to this design decision: #12 (comment). (n = 1 can highlight only the second layer)

So, the workaround would be a bit tricky. Sorry for the inconvenience, I will consider implementing a new feature for this.

library(glmnetUtils)
library(ggfortify)
#> Loading required package: ggplot2

model <- glmnet(medv ~ ., data = MASS::Boston)

g <- autoplot(model, xvar = "lambda")

l <- g$layers[[2]]
g$layers[[2]] <- NULL

library(gghighlight)

g <- g +
  gghighlight(variable == "crim", use_group_by = FALSE)
#> label_key: variable

g$layers[[3]] <- l
g

^{Created on 2018-09-21 by the reprex package (v0.2.1)}

[hoxo-m]

Thank you!
It made me understood that the cause is not in autoplot.
And the workaround was well done for me.

[yutannihilation]

FYI: This is now possible with the dev version.

library(gghighlight)
#> Loading required package: ggplot2
library(glmnetUtils)
library(ggfortify)

model <- glmnet(medv ~ ., data = MASS::Boston)

autoplot(model, xvar = "lambda") +
  gghighlight(variable == "crim", use_group_by = FALSE)
#> Warning: Could not calculate the predicate for layer 2; ignored
#> label_key: variable

^{Created on 2018-12-28 by the reprex package (v0.2.1)}