Experimenting with the opencontext.org API using R

opencontext.r

# To get to animal bones in Turkey: 
q <-   "http://146.148.79.138/sets/Turkey.json?geodeep=10&prop=oc-gen-cat-animal-bone&prop=oc-zoo-anatomical-meas---oc-zoo-von-den-driesch-bone-meas---oc-zoo-bd"

# httr method

library(devtools)
install_github('hadley/httr')
library(httr)
library(jsonlite)

out <- GET(url = q)
# did it work?
http_status(out)
data <- content(out)

# what do we have?
data
length(data)
names(data)

str(data)
str(data$type)
str(data$features)
names(data$features)


data$features[[3]]

# or jsonlite method (don't need httr)

library(jsonlite)
data <- fromJSON(q)
str(data)
data[sapply(data, is.data.frame)]
write.csv(data, "opencontext.csv")



# http://opencontext.org/about/services

###########################################################
# retrieves items located in the region bound by coords
N_min <- 19.382
E_min <- 34.800
N_max <- 29.667
E_max <- 44.000 
query <- paste0("http://opencontext.org/sets/.json?bBox=",E_min, ",", N_min, ",", E_max, ",", N_max)
out <- GET(url = query)
# did it work?
http_status(out)
data <- content(out)
# what do we have?
length(data)
names(data)
data$results[[2]]
data$facets$context
# find the projects within the region
projects <- do.call(rbind.data.frame, data$facets$project)
projects$name
# find the categories of evidence in the region
# and counts of each 
categories <- do.call(rbind.data.frame, data$facets$category)
# find all the records of one category of evidence


animalbone <- data$results[sapply(data$results, function(i) i$category == "Animal Bone")]
do.call(rbind.data.frame, (sapply(animalbone, function(i) i$var_vals))
do.call(rbind.data.frame, animalbone)



##########################################################
# get items by date range
# from Open Context with dates ranging from 1500 BCE to 500 CE.
start <- -1500 # early
end <- 1000 # late
qry <- paste0("http://opencontext.org/sets/.json?t-start=", start, "&t-end=", "end")
out <- GET(url = qry)
http_status(out)
data <- content(out)

# get project names
projs <- (sapply(1:length(data$results), function(i) data$results[[i]]$project))


# get tentative dating
tent_dates <- lapply(1:length(data$results), function(i) data$results[[i]]$var_vals$`Tentative Date`)
# change NULL to NA and keep them when unlist-ing
# http://r.789695.n4.nabble.com/need-help-with-unlist-losing-NULL-values-td4633837.html
tent_dates[unlist(lapply(tent_dates , is.null))] <- NA

tent_dates <- unlist(tent_dates)

data.frame(projs = projs, tent_dates = tent_dates)

Looks good. The new version of the API has a bounding box query too. But it is not documented because because I made a goof translating between GeoJSON ordering of coordinates and Solr's ordering of coordinates, see comments in the commit:

ekansa/open-context-py@df93460

I'm still reindexing all of Open Context to fix this goof. For the next few days (as March 27, 2015), the bounding box queries will only work for the coordinates between 0,0 and 90,90. The new bounding box querying adopts the GeoJSON coordinate pattern (doing the same thing as Pleiades). The order is:

         disc-bbox=X (Longitude -South West), Y (Latitude - South West), X (Longitude - North East), Y (Latitude, North East)

See example.
http://146.148.79.138/sets/?disc-bbox=10,10,45,45