Fixed bug when GEOIDs are ordered differently in result chunks.

[weihuangwong]

When a large number of variables (>24) are passed into get_acs, results sometimes (perhaps rarely) contain errors (perhaps many). This is because get_acs divides the variables into two or more requests, and then binds the returned results together into a dataframe, but GEOIDs may be ordered differently for each set of returned results. For example:

library(tidycensus)

variables <- c("B01001_003", "B01001_004", "B01001_005", "B01001_006", "B01001_007", 
"B01001_008", "B01001_009", "B01001_010", "B01001_011", "B01001_012", 
"B01001_013", "B01001_014", "B01001_015", "B01001_016", "B01001_017", 
"B01001_018", "B01001_019", "B01001_020", "B01001_021", "B01001_022", 
"B01001_023", "B01001_024", "B01001_025", "B01001_026", "B25002_002",
"B03003_003")

test1 <- get_acs(geography = "tract", year = 2017, variables = "B03003_003", 
                 output = "wide", state = 36, geometry = F, key = api_key)

test2 <- get_acs(geography = "tract", year = 2017, variables = variables,
                 output = "wide", state = 36, geometry = F, key = api_key)

identical(test1$B03003_003E[test1$GEOID == "36047037500"], 
          test2$B03003_003E[test2$GEOID == "36047037500"]) # FALSE

The proposed fix uses full_join on GEOID rather than bind_cols.