From 85306007f4f97296f5d48c3edaadb163109c48e1 Mon Sep 17 00:00:00 2001
From: Erik de Vries <erik@devries.pm>
Date: Wed, 16 Jan 2019 19:34:37 +0100
Subject: [PATCH] class_update: added words and clean parameters, in addition
 to text parameter, to be able to set data preprocessing exactly the same as
 in the trained model

---
 R/class_update.R    | 8 +++++---
 man/class_update.Rd | 8 ++++++--
 2 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/R/class_update.R b/R/class_update.R
index 26c5039..d02204f 100644
--- a/R/class_update.R
+++ b/R/class_update.R
@@ -6,7 +6,9 @@
 #' @param model_final The classification model (e.g. output from textstat_nb(), svm() or others)
 #' @param dfm_words A dfm containing all the words and only the words used to generate the model (is used for subsetting)
 #' @param varname String containing the variable name to use for the classification result, usually has the format computerCodes.varname
-#' @param text String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud"
+#' @param words String indicating the number of words to keep from each document (maximum document length), 999 indicates the whole document
+#' @param text String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud", or ud_upos combining lemmas with upos tags
+#' @param clean Boolean indicating whether the results should be cleaned by removing words matching regex (see code).
 #' @param es_super Password for write access to ElasticSearch
 #' @return As this is a nested function used within elasticizer, there is no return output
 #' @export
@@ -15,9 +17,9 @@
 #################################################################################################
 #################################### Update any kind of classification ##########################
 #################################################################################################
-class_update <- function(out, localhost = T, model_final, dfm_words, varname, text, es_super = .rs.askForPassword('ElasticSearch WRITE')) {
+class_update <- function(out, localhost = T, model_final, dfm_words, varname, text, words, clean, es_super = .rs.askForPassword('ElasticSearch WRITE')) {
   print('updating')
-  dfm <- dfm_gen(out, text = text) %>%
+  dfm <- dfm_gen(out, text = text, words = words, clean = clean) %>%
     dfm_keep(dfm_words, valuetype="fixed", verbose=T)
   pred <- data.frame(id = out$`_id`, pred = predict(model_final, newdata = dfm))
   bulk <- apply(pred, 1, bulk_writer, varname = varname, type = 'set')
diff --git a/man/class_update.Rd b/man/class_update.Rd
index ba6224d..d91a80e 100644
--- a/man/class_update.Rd
+++ b/man/class_update.Rd
@@ -5,7 +5,7 @@
 \title{Classifier function for use in combination with the elasticizer function as 'update' parameter (without brackets), see elasticizer documentation for more information}
 \usage{
 class_update(out, localhost = T, model_final, dfm_words, varname, text,
-  es_super = .rs.askForPassword("ElasticSearch WRITE"))
+  words, clean, es_super = .rs.askForPassword("ElasticSearch WRITE"))
 }
 \arguments{
 \item{out}{Does not need to be defined explicitly! (is already parsed in the elasticizer function)}
@@ -18,7 +18,11 @@ class_update(out, localhost = T, model_final, dfm_words, varname, text,
 
 \item{varname}{String containing the variable name to use for the classification result, usually has the format computerCodes.varname}
 
-\item{text}{String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud"}
+\item{text}{String indicating whether the "merged" field will contain the "full" text, old-style "lemmas" (will be deprecated), new-style "ud", or ud_upos combining lemmas with upos tags}
+
+\item{words}{String indicating the number of words to keep from each document (maximum document length), 999 indicates the whole document}
+
+\item{clean}{Boolean indicating whether the results should be cleaned by removing words matching regex (see code).}
 
 \item{es_super}{Password for write access to ElasticSearch}
 }