SCM

SCM Repository

[tm] Annotation of /pkg/R/corpus.R
ViewVC logotype

Annotation of /pkg/R/corpus.R

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1108 - (view) (download)

1 : feinerer 17 # Author: Ingo Feinerer
2 :    
3 : feinerer 985 .PCorpus <- function(x, cmeta, dmeta, dbcontrol) {
4 :     attr(x, "CMetaData") <- cmeta
5 :     attr(x, "DMetaData") <- dmeta
6 :     attr(x, "DBControl") <- dbcontrol
7 :     class(x) <- c("PCorpus", "Corpus", "list")
8 :     x
9 :     }
10 : feinerer 988 DBControl <- function(x) attr(x, "DBControl")
11 : feinerer 985
12 :     PCorpus <- function(x,
13 : feinerer 1073 readerControl = list(reader = x$DefaultReader, language = "en"),
14 : feinerer 946 dbControl = list(dbName = "", dbType = "DB1"),
15 :     ...) {
16 : feinerer 985 readerControl <- prepareReader(readerControl, x$DefaultReader, ...)
17 : feinerer 63
18 : feinerer 946 if (!filehash::dbCreate(dbControl$dbName, dbControl$dbType))
19 :     stop("error in creating database")
20 :     db <- filehash::dbInit(dbControl$dbName, dbControl$dbType)
21 : feinerer 712
22 : feinerer 946 # Allocate memory in advance if length is known
23 : feinerer 1074 tdl <- if (x$Length > 0)
24 : feinerer 985 vector("list", as.integer(x$Length))
25 : feinerer 946 else
26 :     list()
27 : feinerer 869
28 : feinerer 946 counter <- 1
29 : feinerer 985 while (!eoi(x)) {
30 :     x <- stepNext(x)
31 :     elem <- getElem(x)
32 : feinerer 1070 doc <- readerControl$reader(elem, readerControl$language, if (is.null(x$Names)) as.character(counter) else x$Names[counter])
33 : feinerer 946 filehash::dbInsert(db, ID(doc), doc)
34 : feinerer 985 if (x$Length > 0) tdl[[counter]] <- ID(doc)
35 : feinerer 946 else tdl <- c(tdl, ID(doc))
36 :     counter <- counter + 1
37 :     }
38 : feinerer 1064 names(tdl) <- x$Names
39 : feinerer 63
40 : feinerer 946 df <- data.frame(MetaID = rep(0, length(tdl)), stringsAsFactors = FALSE)
41 :     filehash::dbInsert(db, "DMetaData", df)
42 :     dmeta.df <- data.frame(key = "DMetaData", subset = I(list(NA)))
43 : feinerer 712
44 : feinerer 985 .PCorpus(tdl, .MetaDataNode(), dmeta.df, dbControl)
45 :     }
46 : feinerer 71
47 : feinerer 985 .VCorpus <- function(x, cmeta, dmeta) {
48 :     attr(x, "CMetaData") <- cmeta
49 :     attr(x, "DMetaData") <- dmeta
50 :     class(x) <- c("VCorpus", "Corpus", "list")
51 :     x
52 : feinerer 946 }
53 : feinerer 21
54 : feinerer 1021 # Register S3 corpus classes to be recognized by S4 methods. This is
55 :     # mainly a fix to be compatible with packages which were originally
56 :     # developed to cooperate with corresponding S4 tm classes. Necessary
57 :     # since tm's class architecture was changed to S3 since tm version 0.5.
58 :     setOldClass(c("VCorpus", "Corpus", "list"))
59 :    
60 : feinerer 946 # The "..." are additional arguments for the FunctionGenerator reader
61 : feinerer 985 VCorpus <- Corpus <- function(x,
62 : feinerer 1073 readerControl = list(reader = x$DefaultReader, language = "en"),
63 : feinerer 1063 ...) {
64 : feinerer 985 readerControl <- prepareReader(readerControl, x$DefaultReader, ...)
65 : feinerer 49
66 : feinerer 946 # Allocate memory in advance if length is known
67 : feinerer 985 tdl <- if (x$Length > 0)
68 :     vector("list", as.integer(x$Length))
69 : feinerer 946 else
70 :     list()
71 : feinerer 72
72 : feinerer 985 if (x$Vectorized)
73 : feinerer 987 tdl <- mapply(function(x, id) readerControl$reader(x, readerControl$language, id),
74 :     pGetElem(x),
75 : feinerer 1070 id = if (is.null(x$Names)) as.character(seq_len(x$Length)) else x$Names,
76 : feinerer 987 SIMPLIFY = FALSE)
77 : feinerer 946 else {
78 :     counter <- 1
79 : feinerer 985 while (!eoi(x)) {
80 :     x <- stepNext(x)
81 :     elem <- getElem(x)
82 : feinerer 1070 doc <- readerControl$reader(elem, readerControl$language, if (is.null(x$Names)) as.character(counter) else x$Names[counter])
83 : feinerer 985 if (x$Length > 0)
84 : feinerer 946 tdl[[counter]] <- doc
85 :     else
86 :     tdl <- c(tdl, list(doc))
87 :     counter <- counter + 1
88 :     }
89 :     }
90 : feinerer 1063 names(tdl) <- x$Names
91 : feinerer 946 df <- data.frame(MetaID = rep(0, length(tdl)), stringsAsFactors = FALSE)
92 : feinerer 985 .VCorpus(tdl, .MetaDataNode(), df)
93 :     }
94 : feinerer 72
95 : feinerer 985 `[.PCorpus` <- function(x, i) {
96 :     if (missing(i)) return(x)
97 :     index <- attr(x, "DMetaData")[[1 , "subset"]]
98 :     attr(x, "DMetaData")[[1 , "subset"]] <- if (is.numeric(index)) index[i] else i
99 :     dmeta <- attr(x, "DMetaData")
100 : feinerer 995 .PCorpus(NextMethod("["), CMetaData(x), dmeta, DBControl(x))
101 : feinerer 946 }
102 : feinerer 72
103 : feinerer 985 `[.VCorpus` <- function(x, i) {
104 :     if (missing(i)) return(x)
105 : feinerer 995 .VCorpus(NextMethod("["), CMetaData(x), DMetaData(x)[i, , drop = FALSE])
106 : feinerer 985 }
107 : feinerer 49
108 : feinerer 985 `[<-.PCorpus` <- function(x, i, value) {
109 :     db <- filehash::dbInit(DBControl(x)[["dbName"]], DBControl(x)[["dbType"]])
110 :     counter <- 1
111 :     for (id in unclass(x)[i]) {
112 : feinerer 1025 if (identical(length(value), 1L)) db[[id]] <- value
113 : feinerer 985 else db[[id]] <- value[[counter]]
114 :     counter <- counter + 1
115 : feinerer 829 }
116 : feinerer 985 x
117 : feinerer 828 }
118 :    
119 : feinerer 1108 .map_name_index <- function(x, i) {
120 :     if (is.character(i)) {
121 :     if (is.null(names(x)))
122 :     match(i, meta(x, "ID", type = "local"))
123 :     else
124 :     match(i, names(x))
125 :     }
126 :     i
127 :     }
128 :    
129 : feinerer 985 `[[.PCorpus` <- function(x, i) {
130 : feinerer 1108 i <- .map_name_index(x, i)
131 : feinerer 985 db <- filehash::dbInit(DBControl(x)[["dbName"]], DBControl(x)[["dbType"]])
132 : feinerer 995 filehash::dbFetch(db, NextMethod("[["))
133 : feinerer 985 }
134 :     `[[.VCorpus` <- function(x, i) {
135 : feinerer 1108 i <- .map_name_index(x, i)
136 : feinerer 985 lazyTmMap <- meta(x, tag = "lazyTmMap", type = "corpus")
137 :     if (!is.null(lazyTmMap))
138 :     .Call("copyCorpus", x, materialize(x, i))
139 : feinerer 995 NextMethod("[[")
140 : feinerer 985 }
141 : feinerer 886
142 : feinerer 985 `[[<-.PCorpus` <- function(x, i, value) {
143 : feinerer 1108 i <- .map_name_index(x, i)
144 : feinerer 985 db <- filehash::dbInit(DBControl(x)[["dbName"]], DBControl(x)[["dbType"]])
145 :     index <- unclass(x)[[i]]
146 :     db[[index]] <- value
147 :     x
148 : feinerer 946 }
149 : feinerer 985 `[[<-.VCorpus` <- function(x, i, value) {
150 : feinerer 1108 i <- .map_name_index(x, i)
151 : feinerer 985 # Mark new objects as not active for lazy mapping
152 :     lazyTmMap <- meta(x, tag = "lazyTmMap", type = "corpus")
153 :     if (!is.null(lazyTmMap)) {
154 :     lazyTmMap$index[i] <- FALSE
155 :     meta(x, tag = "lazyTmMap", type = "corpus") <- lazyTmMap
156 :     }
157 :     # Set the value
158 :     cl <- class(x)
159 : feinerer 995 y <- NextMethod("[[<-")
160 :     class(y) <- cl
161 :     y
162 : feinerer 985 }
163 : feinerer 946
164 : feinerer 1004 # Update NodeIDs of a CMetaData tree
165 :     .update_id <- function(x, id = 0, mapping = NULL, left.mapping = NULL, level = 0) {
166 :     # Traversal of (binary) CMetaData tree with setup of NodeIDs
167 : feinerer 985 set_id <- function(x) {
168 : feinerer 988 x$NodeID <- id
169 : feinerer 697 id <<- id + 1
170 :     level <<- level + 1
171 : feinerer 988 if (length(x$Children) > 0) {
172 :     mapping <<- cbind(mapping, c(x$Children[[1]]$NodeID, id))
173 :     left <- set_id(x$Children[[1]])
174 : feinerer 697 if (level == 1) {
175 :     left.mapping <<- mapping
176 :     mapping <<- NULL
177 :     }
178 : feinerer 988 mapping <<- cbind(mapping, c(x$Children[[2]]$NodeID, id))
179 :     right <- set_id(x$Children[[2]])
180 : feinerer 71
181 : feinerer 988 x$Children <- list(left, right)
182 : feinerer 71 }
183 : feinerer 697 level <<- level - 1
184 : feinerer 985 x
185 : feinerer 71 }
186 : feinerer 985 list(root = set_id(x), left.mapping = left.mapping, right.mapping = mapping)
187 : feinerer 71 }
188 :    
189 : feinerer 1004 # Find indices to be updated for a CMetaData tree
190 :     .find_indices <- function(x) {
191 :     indices.mapping <- NULL
192 :     for (m in levels(as.factor(DMetaData(x)$MetaID))) {
193 :     indices <- (DMetaData(x)$MetaID == m)
194 :     indices.mapping <- c(indices.mapping, list(m = indices))
195 :     names(indices.mapping)[length(indices.mapping)] <- m
196 :     }
197 :     indices.mapping
198 :     }
199 :    
200 : feinerer 985 c2 <- function(x, y, ...) {
201 :     # Update the CMetaData tree
202 :     cmeta <- .MetaDataNode(0, list(merge_date = as.POSIXlt(Sys.time(), tz = "GMT"), merger = Sys.getenv("LOGNAME")), list(CMetaData(x), CMetaData(y)))
203 : feinerer 1004 update.struct <- .update_id(cmeta)
204 : feinerer 71
205 : feinerer 985 new <- .VCorpus(c(unclass(x), unclass(y)), update.struct$root, NULL)
206 : feinerer 720
207 : feinerer 985 # Find indices to be updated for the left tree
208 : feinerer 1004 indices.mapping <- .find_indices(x)
209 : feinerer 958
210 : feinerer 985 # Update the DMetaData data frames for the left tree
211 :     for (i in 1:ncol(update.struct$left.mapping)) {
212 :     map <- update.struct$left.mapping[,i]
213 :     DMetaData(x)$MetaID <- replace(DMetaData(x)$MetaID, indices.mapping[[as.character(map[1])]], map[2])
214 :     }
215 : feinerer 689
216 : feinerer 985 # Find indices to be updated for the right tree
217 : feinerer 1004 indices.mapping <- .find_indices(y)
218 : feinerer 71
219 : feinerer 985 # Update the DMetaData data frames for the right tree
220 :     for (i in 1:ncol(update.struct$right.mapping)) {
221 :     map <- update.struct$right.mapping[,i]
222 :     DMetaData(y)$MetaID <- replace(DMetaData(y)$MetaID, indices.mapping[[as.character(map[1])]], map[2])
223 :     }
224 : feinerer 71
225 : feinerer 985 # Merge the DMetaData data frames
226 :     labels <- setdiff(names(DMetaData(y)), names(DMetaData(x)))
227 :     na.matrix <- matrix(NA, nrow = nrow(DMetaData(x)), ncol = length(labels), dimnames = list(row.names(DMetaData(x)), labels))
228 :     x.dmeta.aug <- cbind(DMetaData(x), na.matrix)
229 :     labels <- setdiff(names(DMetaData(x)), names(DMetaData(y)))
230 :     na.matrix <- matrix(NA, nrow = nrow(DMetaData(y)), ncol = length(labels), dimnames = list(row.names(DMetaData(y)), labels))
231 :     y.dmeta.aug <- cbind(DMetaData(y), na.matrix)
232 :     DMetaData(new) <- rbind(x.dmeta.aug, y.dmeta.aug)
233 : feinerer 71
234 : feinerer 985 new
235 :     }
236 : feinerer 71
237 : feinerer 985 c.Corpus <-
238 :     function(x, ..., recursive = FALSE)
239 :     {
240 :     args <- list(...)
241 : feinerer 71
242 : feinerer 1025 if (identical(length(args), 0L))
243 : feinerer 985 return(x)
244 : feinerer 71
245 : feinerer 985 if (!all(unlist(lapply(args, inherits, class(x)))))
246 :     stop("not all arguments are of the same corpus type")
247 : feinerer 71
248 : feinerer 985 if (inherits(x, "PCorpus"))
249 :     stop("concatenation of corpora with underlying databases is not supported")
250 : feinerer 689
251 : feinerer 1095 l <- base::c(list(x), args)
252 :     if (recursive)
253 :     Reduce(c2, l)
254 :     else {
255 :     l <- do.call("c", lapply(l, unclass))
256 :     .VCorpus(l,
257 :     cmeta = .MetaDataNode(),
258 :     dmeta = data.frame(MetaID = rep(0, length(l)), stringsAsFactors = FALSE))
259 :     }
260 : feinerer 985 }
261 : feinerer 54
262 : feinerer 985 c.TextDocument <- function(x, ..., recursive = FALSE) {
263 :     args <- list(...)
264 : feinerer 72
265 : feinerer 1025 if (identical(length(args), 0L))
266 : feinerer 985 return(x)
267 : feinerer 72
268 : feinerer 985 if (!all(unlist(lapply(args, inherits, class(x)))))
269 :     stop("not all arguments are text documents")
270 : feinerer 54
271 : feinerer 985 dmeta <- data.frame(MetaID = rep(0, length(list(x, ...))), stringsAsFactors = FALSE)
272 :     .VCorpus(list(x, ...), .MetaDataNode(), dmeta)
273 :     }
274 : feinerer 55
275 : feinerer 985 print.Corpus <- function(x, ...) {
276 :     cat(sprintf(ngettext(length(x),
277 :     "A corpus with %d text document\n",
278 :     "A corpus with %d text documents\n"),
279 :     length(x)))
280 :     invisible(x)
281 :     }
282 :    
283 : feinerer 988 summary.Corpus <- function(object, ...) {
284 :     print(object)
285 :     if (length(DMetaData(object)) > 0) {
286 :     cat(sprintf(ngettext(length(attr(CMetaData(object), "MetaData")),
287 : feinerer 985 "\nThe metadata consists of %d tag-value pair and a data frame\n",
288 :     "\nThe metadata consists of %d tag-value pairs and a data frame\n"),
289 : feinerer 988 length(CMetaData(object)$MetaData)))
290 : feinerer 985 cat("Available tags are:\n")
291 : feinerer 988 cat(strwrap(paste(names(CMetaData(object)$MetaData), collapse = " "), indent = 2, exdent = 2), "\n")
292 : feinerer 985 cat("Available variables in the data frame are:\n")
293 : feinerer 988 cat(strwrap(paste(names(DMetaData(object)), collapse = " "), indent = 2, exdent = 2), "\n")
294 : feinerer 985 }
295 :     }
296 :    
297 : feinerer 938 inspect <- function(x) UseMethod("inspect", x)
298 : feinerer 946 inspect.PCorpus <- function(x) {
299 : feinerer 938 summary(x)
300 :     cat("\n")
301 : feinerer 946 db <- filehash::dbInit(DBControl(x)[["dbName"]], DBControl(x)[["dbType"]])
302 :     show(filehash::dbMultiFetch(db, unlist(x)))
303 : feinerer 938 }
304 : feinerer 963 inspect.VCorpus <- function(x) {
305 : feinerer 946 summary(x)
306 :     cat("\n")
307 :     print(noquote(lapply(x, identity)))
308 :     }
309 : feinerer 65
310 : feinerer 985 lapply.PCorpus <- function(X, FUN, ...) {
311 :     db <- filehash::dbInit(DBControl(X)[["dbName"]], DBControl(X)[["dbType"]])
312 :     lapply(filehash::dbMultiFetch(db, unlist(X)), FUN, ...)
313 :     }
314 :     lapply.VCorpus <- function(X, FUN, ...) {
315 :     lazyTmMap <- meta(X, tag = "lazyTmMap", type = "corpus")
316 :     if (!is.null(lazyTmMap))
317 :     .Call("copyCorpus", X, materialize(X))
318 :     base::lapply(X, FUN, ...)
319 :     }
320 : feinerer 719
321 : feinerer 985 writeCorpus <- function(x, path = ".", filenames = NULL) {
322 :     filenames <- file.path(path,
323 :     if (is.null(filenames)) unlist(lapply(x, function(x) sprintf("%s.txt", ID(x))))
324 :     else filenames)
325 :     i <- 1
326 :     for (o in x) {
327 :     writeLines(as.PlainTextDocument(o), filenames[i])
328 :     i <- i + 1
329 : feinerer 836 }
330 : feinerer 985 }

R-Forge@R-project.org
ViewVC Help
Powered by ViewVC 1.0.0  
Thanks to:
Vienna University of Economics and Business University of Wisconsin - Madison Powered By FusionForge