mutate_geocode() is incompatible with the behavior of tbl_df-type data frames

[dannguyen]

If the data frame that is passed into mutate_geocode() happens to be a class of tbl_df (via dplyr, an error will arise due to tbl_df's different subsetting behavior.

• The issue is described here: Problem with table_df
• At this point, Hadley believes that tbl_df does not need to change.

Demonstration of error:

library(ggmap)
df <- data.frame(
  address = c("1600 Pennsylvania Avenue, Washington DC", "", "houston texas"),
  stringsAsFactors = FALSE
)
## Convert to tbl_df
tdf <- as_data_frame(df)
mutate_geocode(tdf, location = address, source = "google")

Error message:

  Error: is.character(location) is not TRUE
  In addition: Warning message:
  drop ignored 

I'm not sure of how often ggmap users will have a situation in which their data frames are actually tbl_dfs...all I know is that it happened to me on my very first spin of mutate_geocode :).

In the pull request I've made what I think fixes the issue without requiring the user to convert their tbl_df before running mutate_geocode(). I'm not sure if this works for all cases or introduces any other kind of breakage. Or if any of ggmap's other vector functions might need the same fix.

mutate_geocode() coerces the result into a data.frame...not sure if that should be changed in case a user is expecting tbl_df() back...

[crazybilly]

With regard to coercing the results to data.frame, ggmap is already importing dplyr::bind_cols. Couldn't we change the last line to the following to maintain the class of the original data frame?

bind_cols(data, gcdf)

[mine-cetinkaya-rundel]

Are there plans to merge this, or a similar fix that addresses the same issue?