fix errors with ncbi_byname reported in #126

R/ncbi_byname.R

@@ -20,8 +20,7 @@
 #' species <- c("Colletes similis","Halictus ligatus","Perdita californica")
 #' ncbi_byname(taxa=species, gene = c("coi", "co1"), seqrange = "1:2000")
 #' }
-ncbi_byname <- function(taxa, gene="COI", seqrange="1:3000", getrelated=FALSE,
-                        verbose=TRUE, ...) {
+ncbi_byname <- function(taxa, gene="COI", seqrange="1:3000", getrelated=FALSE, verbose=TRUE, batch_size = 100, ...) {
   foo <- function(xx) {
     mssg(verbose, paste("Working on ", xx, "...", sep = ""))
     mssg(verbose, "...retrieving sequence IDs...")
@@ -51,9 +50,10 @@ ncbi_byname <- function(taxa, gene="COI", seqrange="1:3000", getrelated=FALSE,
       if (!getrelated) {
         mssg(verbose, paste("no sequences of ", gene, " for ", xx, sep = ""))
         res <- data.frame(
-          xx, NA_character_, NA_real_, NA_character_, NA_real_, NA_character_, NA_character_,
+          taxon = xx, gene_desc = NA_character_, 
+          gi_no = NA_real_, acc_no = NA_character_, length = NA_real_, 
+          sequence = NA_character_, spused = NA_character_,
           stringsAsFactors = FALSE)
-        names(res) <- NULL
       } else {
         mssg(verbose, "...retrieving sequence IDs for related species...")
         newname <- strsplit(xx, " ")[[1]][[1]]
@@ -68,9 +68,10 @@ ncbi_byname <- function(taxa, gene="COI", seqrange="1:3000", getrelated=FALSE,
         if (as.numeric(xml2::xml_text(xml2::xml_find_all(out, "//Count")[[1]])) == 0) {
           mssg(verbose, paste("no sequences of ", gene, " for ", xx, " or ", newname, sep = ""))
           res <- data.frame(
-            xx, NA_character_, NA_real_, NA_character_, NA_real_, NA_character_, NA_character_,
+            taxon = xx, gene_desc = NA_character_, 
+            gi_no = NA_real_, acc_no = NA_character_, length = NA_real_, 
+            sequence = NA_character_, spused = NA_character_,
             stringsAsFactors = FALSE)
-          names(res) <- NULL
         } else {
           ## For each species = get GI number with longest sequence
           mssg(verbose, "...retrieving sequence ID with longest sequence length...")
@@ -88,30 +89,53 @@ ncbi_byname <- function(taxa, gene="COI", seqrange="1:3000", getrelated=FALSE,
       ## For each species = get GI number with longest sequence
       mssg(verbose, "...retrieving sequence ID with longest sequence length...")
       # construct query for species
-      querysum <- list(db = "nucleotide", api_key = ncbi_key(),
-        id = paste(make_ids(out), collapse = " "))
-      z <- con$get("entrez/eutils/esummary.fcgi", query = querysum)
-      txt <- z$parse("UTF-8")
-      res <- parse_ncbi(xx, xml2::xml_find_all(xml2::read_xml(txt), "//eSummaryResult")[[1]], verbose)
+      ids <- make_ids(out)
+      res_list <- lapply(seq(1, length(ids), by = batch_size), function(i) {
+        querysum <- list(db = "nucleotide", 
+                         id = paste(ids[i:min(i + batch_size - 1, length(ids))], collapse = " "),
+                         api_key = ncbi_key())
+        z <- con$get("entrez/eutils/esummary.fcgi", query = querysum)
+        z$raise_for_status()
+        xml2::xml_find_all(xml2::read_xml(z$parse("UTF-8")), "//eSummaryResult")
+      })
+
+      res <- do.call(rbind, lapply(res_list, function(x) parse_ncbi(xx, x, verbose)))
     }
-
+    
     mssg(verbose, "...done.")
     stats::setNames(res, c("taxon", "gene_desc", "gi_no", "acc_no", "length", "sequence", "spused"))
   }
-
+  
   foo_safe <- tryfail(NULL, foo)
-  if (length(taxa) == 1){ foo_safe(taxa) } else { lapply(taxa, foo_safe) }
+  if (length(taxa) == 1){ 
+    return(foo_safe(taxa))
+  } else { 
+    return(do.call(rbind, lapply(taxa, foo_safe))) 
+  }
 }
 
 parse_ncbi <- function(xx, z, verbose) {
+
   names <- xml2::xml_attr(xml2::xml_find_all(z, "//Item"), "Name") # gets names of values in summary
+
+  if(length(names) == 0) {
+    message("No sequences found for ", xx)
+    return(data.frame(taxon = xx, gene_desc = NA, 
+                      gi_no = NA, acc_no = NA, length = NA, 
+                      sequence = NA, spused = NA, 
+                      stringsAsFactors = FALSE))
+  }
+
   prd <- xml2::xml_text(xml2::xml_find_all(z, '//Item[@Name="Caption"]')) #  get access numbers
   prd <- sapply(prd, function(x) strsplit(x, "_")[[1]][[1]], USE.NAMES=FALSE)
   l_ <- as.numeric(xml2::xml_text(xml2::xml_find_all(z, '//Item[@Name="Length"]'))) # gets seq lengths
   gis <- as.numeric(xml2::xml_text(xml2::xml_find_all(z, '//Item[@Name="Gi"]'))) # gets GI numbers
   sns <- xml2::xml_text(xml2::xml_find_all(z, '//Item[@Name="Title"]')) # gets seq lengths # get spp names
   df <- data.frame(gis=gis, length=l_, spnames=sns, predicted=prd, stringsAsFactors = FALSE)
   df <- df[!df$predicted %in% c("XM","XR"),] # remove predicted sequences
+  if (nrow(df) == 0) {
+    return(data.frame(taxon = xx, gene_desc = NA, gi_no = NA, acc_no = NA, length = NA, sequence = NA, spused = NA, stringsAsFactors = FALSE))
+  }
   gisuse <- df[which.max(x=df$length),] # picks longest sequnence length
   if (NROW(gisuse) > 1) {
     gisuse <- gisuse[sample(NROW(gisuse), 1), ]
@@ -126,12 +150,25 @@ parse_ncbi <- function(xx, z, verbose) {
   outseq <- w$parse("UTF-8")
   seq <- gsub("\n", "", strsplit(sub("\n", "<<<", outseq), "<<<")[[1]][[2]])
   accessnum <- strsplit(outseq, "\\|")[[1]][4]
-  outt <- list(xx, as.character(gisuse[,3]), gisuse[,1], accessnum, gisuse[,2], seq)
+  outt <- list(taxon = xx, gene_desc = as.character(gisuse[,3]), 
+               gi_no = gisuse[,1], acc_no = accessnum, 
+               length = gisuse[,2], sequence = seq, 
+               spused = paste(
+                 strsplit(as.character(gisuse[,3]), 
+                          " ")[[1]][1:2], 
+                 collapse = " "))
 
   spused <- paste(strsplit(outt[[2]], " ")[[1]][1:2], sep = "", collapse = " ")
-  outoutout <- data.frame(outt, spused = spused, stringsAsFactors = FALSE)
-  names(outoutout) <- NULL
-  outoutout
+  outoutout <- data.frame(outt, stringsAsFactors = FALSE)
+
+  return(outoutout)
 }
 
 make_ids <- function(x) as.numeric(xml2::xml_text(xml2::xml_find_all(x, "//IdList//Id")))
+
+tryfail <- function(default, code) {
+  tryCatch(code, error = function(e) { 
+    message(e)
+    return(default) 
+  })
+}

tests/testthat/test-ncbi.R

@@ -0,0 +1,55 @@
+context("NCBI tests")
+
+test_one_species <- "Acer rubrum"
+test_three_species <- c("Colletes similis", "Halictus ligatus", "Perdita californica")
+test_genes <- c("coi", "co1")
+test_that("ncbi_byname gets seq for single species", {
+  skip_on_cran()
+  result <- ncbi_byname("Acer rubrum")
+  expect_equal(ncol(result), 7)
+  expect_equal(nrow(result), 1)
+  expect_true(!is.na(result$taxon))
+  expect_true(!is.na(result$sequence))
+})
+
+test_that("ncbi_byname gets seq for multiple species", {
+  skip_on_cran()
+  result <- ncbi_byname(test_three_species)
+  expect_equal(ncol(result), 7)
+  expect_gte(nrow(result), 3)
+  expect_true(all(!is.na(result$taxon)))
+  expect_true(all(!is.na(result$sequence)))
+})
+
+test_that("ncbi_byname handles cases with no results"), {
+  skip_on_cran()
+  result <- ncbi_byname("This is not a species")
+  expect_equal(ncol(result), 7)
+  expect_equal(nrow(result), 1)
+  expect_true(all(is.na(result[,2:7])))
+})
+
+test_that("ncbi_byname gets seq for single gene", {
+  skip_on_cran()
+  result <- ncbi_byname(test_one_species, gene = "coi")
+  expect_equal(ncol(result), 7)
+  expect_gte(nrow(result), 1)
+  expect_true(all(!is.na(result$taxon)))
+  expect_true(all(!is.na(result$sequence)))
+})
+
+## regression tests for https://github.com/ropensci/traits/issues/126
+
+test_that("ncbi_byname works under subset of conditions from issue #126", {
+  skip_on_cran()
+
+  res <- ncbi_byname(taxa = "Coryphaena hippurus", gene = c("Coi"), seqrange = "1:2000")
+  expect_true(is.data.frame(res))
+  res <- ncbi_byname(taxa = "Coryphaena hippurus", gene = c("Coi"), seqrange = "1:2000")
+  res2 <- ncbi_byname(taxa = "Coryphaena hippurus", gene = c("Coi"), seqrange = "500:750")
+  expect_true(is.data.frame(res2))
+
+  res3 <- ncbi_byname(taxa = "Sardinops melanostictus", gene = c("12s"), seqrange = "1:2000"))
+  expect_true(is.data.frame(res3))
+
+})