Non-ASCII characters lead to NA geocoding records

[dvmasterov]

Maybe this is obvious, but I am seeing some strange behavior with non-ASCII characters on Mac OS. Here's my version info:

> packageVersion("censusxy")
[1] ‘1.0.0’
> getRversion()
[1] ‘4.0.2’

library(censusxy)
library(dplyr)
library(stringi)
library(sf)

> # this works (as does the web ui)
> g<-cxy_single('412 45th Strèet','Oakland','CA','94609', return = 'geographies', vintage = 'Current_Current')
> summary(as.factor(g$cxy_status))
integer(0)
> 
> 
> # this breaks
> my_df <- data.frame(street= '412 45th Strèet', city = 'Oakland', state='CA', zip ='94609')
> geocoded_data <- cxy_geocode(my_df,
+                                     street = "street", 
+                                     city = "city", 
+                                     state = "state", 
+                                     zip = "zip",
+                                     output = "full", 
+                                     class = "dataframe", 
+                                     return="geographies",
+                                     vintage ='Current_Current')
> 
> summary(as.factor(geocoded_data$cxy_status))
NA's 
   1 
> 
> my_df <- my_df %>% 
+   mutate_if(is.character,
+             stri_trans_general,
+             id = "latin-ascii")
> 
> # this breaks
> geocoded_data <- cxy_geocode(my_df,
+                              street = "street", 
+                              city = "city", 
+                              state = "state", 
+                              zip = "zip",
+                              output = "full", 
+                              class = "dataframe", 
+                              return="geographies",
+                              vintage ='Current_Current')
> 
> summary(as.factor(geocoded_data$cxy_status))
Match 
    1 

[chris-prener]

This is an issue with the Census Bureau side, and not the package itself. My guess is they can't handle non-ASCII characters. If you can transform them as you've done, I would recommend doing that.

[dvmasterov]

I think a warning or at least a mention in the manual of these requirements would be a nice add. This is not well-documented on the census side (as far as I found). I agree that this is on their end, but it would be nice to spare future users this headache.

[chris-prener]

Yeah, we can make note of it somehow... do you have some "real world" examples of streets with non ASCII characters I could use?