Parse Dataframe string column into Street, City, State & Zip code - sql

Question

I am trying to break up the following fixed string into several columns as street ,city, state & zip code. Is it possible to do this in SQLDF via the INSTR & Subtr method?
Sample Address String. The difficult part is the NV and zip code parsing.
727 Wright Brothers Ln, Las Vegas, NV 89119, USA
I am able to parse the city/street information using sqldf/instr but unable to parse the final two values for state/zip code
parsed_tweetAddressdf <- sqldf("SELECT lon, lat, result, substr(result,0,instr(result,',')) AS street, substr(result,instr(result,',')+1,instr(result,',')-1) AS city from tweetAddressdf")

Answer 1

Here are some alternatives. They all use instr and substr as required by the question although the third also writes out the data and reads it back in (in addition to using instr and substr). Notes at the end point out that it is also easy to do this in plain R or using read.pattern in gsubfn.
1) Assume state, zip and country fields are fixed width With only one sample record it is impossible to know what your general case is but if we assume that every record ends in SS ZZZZZ, USA where SS is the two letter state abbreviation and ZZZZZ is a 5 digit zip then this works:
DF <- data.frame(v = "727 Wright Brothers Ln, Las Vegas, NV 89119, USA")
library(sqldf)
sqldf("select
substr(v, 0, instr(v, ',')) street,
substr(v, instr(v, ',') + 2, length(v) - 16 - instr(v, ',')) city,
substr(v, -13, 2) state,
substr(v, -10, 5) zip
from DF")
giving:
street city state zip
1 727 Wright Brothers Ln Las Vegas NV 89119
2) Separate strictly based on commas (except state/zip) This approach avoids certain assumptions in (1) at the expense of additional complication. It takes the first two comma separated fields, the 2 character state and everything after that to the next comma as the zip.
It uses a triple nested select. The innermost select denoted a parses the input string into: street and a.rest. The next one proceeding outward denoted b returns the street already parsed from a, and parses a.rest into city and the b.rest. The outermost one returns the street and city already parsed plus it returns the two state characters in b.rest and everything beyond them in b.rest to the next comma as zip.
library(sqldf)
sqldf("
select
street,
city,
substr(b.rest, 1, 2) state,
substr(b.rest, 4, instr(b.rest, ',') - 4) zip
from (
select
street,
substr(a.rest, 0, instr(a.rest, ',')) city,
substr(a.rest, instr(a.rest, ',') + 2) rest
from (select
substr(v, 0, instr(v, ',')) street,
substr(v, instr(v, ',') + 2) rest
from DF) a) b
")
giving:
street city state zip
1 727 Wright Brothers Ln Las Vegas NV 89119
3) read.csv.sql If it's OK to write it out and read it back in then we can use read.csv.sql, a wrapper around sqldf. Although the question did not ask for it, this one also parses out the country:
write.table(DF, "addresses.csv", row.names = FALSE, col.names = FALSE,
sep = ",", quote = FALSE)
read.csv.sql("addresses.csv", header = FALSE, sql =
"select V1 street,
V2 city,
substr(V3, 2, 2) state,
substr(V3, 4) zip,
V4 country
from file")
giving:
street city state zip country
1 727 Wright Brothers Ln Las Vegas NV 89119 USA
Note 1: This is also easy in plain R.
dd <- read.table(text = as.character(DF$v), sep = ",",
col.names = c("street", "city", "state_zip", "country"))
transform(dd,
state = substring(state_zip, 2, 3),
zip = substring(state_zip, 4))[c(1, 2, 5, 6, 4)]
giving:
street city state zip country
1 727 Wright Brothers Ln Las Vegas NV 89119 USA
Note 2: It is even easier using read.pattern from gsubfn:
library(gsubfn)
pat <- "(.*), (.*), (..) (.*), (.*)"
read.pattern(text = as.character(DF$v), pattern = pat,
col.names = c("street", "city", "state", "zip", "country"))
giving:
street city state zip country
1 727 Wright Brothers Ln Las Vegas NV 89119 USA