Merge branch 'master' into add-test-adjust

R/lda.R

@@ -54,7 +54,8 @@
 #'   \item{k}{the number of topics.}
 #'   \item{last_iter}{the number of iterations in Gibbs sampling.}
 #'   \item{max_iter}{the maximum number of iterations in Gibbs sampling.}
-#'   \item{auto_iter}{`auto_iter` is used if `TRUE`.}
+#'   \item{auto_iter}{the use of `auto_iter`}
+#'   \item{adjust_alpha}{the value of `adjust_alpha`.}
 #'   \item{alpha}{the smoothing parameter for `theta`.}
 #'   \item{beta}{the smoothing parameter for `phi`.}
 #'   \item{epsilon}{the amount of adjustment for `adjust_alpha`.}

R/seededlda.R

@@ -189,6 +189,7 @@ tfm <- function(x, dictionary, levels = 1,
     if (!quanteda::is.dictionary(dictionary))
         stop("dictionary must be a dictionary object", call. = FALSE)
 
+    docvars(x) <- NULL # sanitize dfm
     dict <- flatten_dictionary(dictionary, levels)
     key <- names(dict)
     feat <- featnames(x)

src/lda.cpp

@@ -1,3 +1,4 @@
+#include <RcppArmadillo.h>
 #include "lib.h"
 #include "dev.h"
 #include "lda.h"
@@ -38,6 +39,7 @@ List cpp_lda(arma::sp_mat &mt, int k, int max_iter, double min_delta,
                         Rcpp::Named("max_iter") = lda.max_iter,
                         Rcpp::Named("last_iter") = lda.iter,
                         Rcpp::Named("auto_iter") = (lda.min_delta == 0),
+                        Rcpp::Named("adjust_alpha") = lda.adjust,
                         Rcpp::Named("alpha") = as<NumericVector>(wrap(lda.alpha)),
                         Rcpp::Named("beta") = as<NumericVector>(wrap(lda.beta)),
                         Rcpp::Named("epsilon") = as<NumericVector>(wrap(lda.epsilon)),

tests/testthat/test-internal.R

@@ -116,6 +116,24 @@ test_that("tfm works with ngrams", {
                  c("un" = 2, "icc" = 2, "other" = 0))
 })
 
+test_that("tfm works with dfm with x in docvars (#87)", {
+
+	dict <- dictionary(list("A" = "a", "B" = "b"))
+	dat <- data.frame(text = c("a b c", "A B C"),
+					  x = c(1, 2))
+	corp <- corpus(dat)
+	toks <- tokens(corp)
+	dfmt <- dfm(toks)
+
+	expect_equal(
+		as.matrix(seededlda:::tfm(dfmt, dict, residula = 1)),
+		matrix(c(2, 0, 0, 0, 2, 0, 0, 0 ,0), nrow = 3,
+			   dimnames = list(c("A", "B", "other"), c("a", "b", "c")))
+	)
+
+})
+
+
 test_that("levels is working", {
 
 	dict <- dictionary(list(A = list(

tests/testthat/test-textmodel_lda.R

@@ -72,7 +72,8 @@ test_that("LDA is working", {
     )
     expect_equal(
         names(lda),
-        c("k", "max_iter", "last_iter", "auto_iter", "alpha", "beta", "epsilon", "gamma",
+        c("k", "max_iter", "last_iter", "auto_iter", "adjust_alpha",
+          "alpha", "beta", "epsilon", "gamma",
           "phi", "theta", "words", "data", "batch_size", "call", "version")
     )
     expect_equal(lda$last_iter, 200)
@@ -124,6 +125,7 @@ test_that("adjust_alpha works", {
 	expect_equivalent(rowSums(lda$theta), rep(1.0, ndoc(lda$data)))
 	expect_equivalent(rowSums(lda$phi), rep(1.0, lda$k))
 
+	expect_equal(lda$adjust_alpha, 0.5)
 	expect_true(all(lda$alpha != 0.5))
 	expect_true(all(lda$epsilon > 0))
 

tests/testthat/test-textmodel_seededlda.R

@@ -104,8 +104,9 @@ test_that("seeded LDA is working", {
     )
     expect_equal(
         names(lda),
-        c("k", "max_iter", "last_iter", "auto_iter", "alpha", "beta", "epsilon", "gamma",
-          "phi", "theta", "words", "data", "batch_size", "call", "version",
+        c("k", "max_iter", "last_iter", "auto_iter", "adjust_alpha",
+          "alpha", "beta", "epsilon", "gamma", "phi", "theta",
+          "words", "data", "batch_size", "call", "version",
           "dictionary", "valuetype", "case_insensitive", "seeds",
           "residual", "weight")
     )