SCM

SCM Repository

[inlinedocs] Annotation of /pkg/inlinedocs/R/parsers.R
ViewVC logotype

Annotation of /pkg/inlinedocs/R/parsers.R

Parent Directory Parent Directory | Revision Log Revision Log


Revision 335 - (view) (download)

1 : tdhock 335 do.not.generate <- function
2 :     ### Make a Parser Function used to indicate that certain Rd files
3 :     ### should not be generated.
4 :     (...
5 :     ### Character strings indicating Rd files without the .Rd suffix.
6 :     ){
7 :     filenames <- c(...)
8 :     function(docs,...){
9 :     for(fn in filenames){
10 :     docs[[fn]] <- list()
11 :     }
12 :     docs$.overwrite <- TRUE
13 :     docs
14 :     }
15 :     ### A Parser Function that will delete items from the outer
16 :     ### Documentation List.
17 :     }
18 :    
19 : tdhock 110 ### combine lists or character strings
20 :     combine <- function(x,y)UseMethod("combine")
21 :    
22 :     ### combine character strings by pasting them together
23 :     combine.character <- function(x,y)
24 :     paste(x,y,sep="\n")
25 :    
26 :     ### combine lists by adding elements or adding to existing elements
27 :     combine.list <- function(x,y){
28 : tdhock 200 toadd <- if(".overwrite"%in%names(y)){
29 :     y <- y[names(y)!=".overwrite"]
30 :     rep(TRUE,length(y))
31 :     }else{
32 :     !names(y)%in%names(x)
33 :     }
34 : tdhock 185 toup <- names(y)[!toadd]
35 : tdhock 110 x[names(y)[toadd]] <- y[toadd]
36 :     for(up in toup)x[[up]] <- combine(x[[up]],y[[up]])
37 : tdhock 200 x
38 : tdhock 110 ### A list, same type as x, but with added elements from y.
39 :     }
40 :    
41 : tdhock 302
42 :     getSource <- function
43 :     ### Extract a function's source code.
44 :     (fun.obj
45 :     ### A function.
46 :     ) {
47 :     srcref <- attr(fun.obj, "srcref")
48 : tdhock 308 if (!is.null(srcref)) {
49 :     ##unlist(strsplit(as.character(srcref), "\n"))
50 :     as.character(srcref)
51 :     }
52 : tdhock 302 else attr(fun.obj, "source")
53 :     ### Source code lines as a character vector.
54 :     }
55 :    
56 : tdhock 110 ### Prefix for code comments used with grep and gsub.
57 : tdhock 257 prefix <- "^[ \t]*###[ \t]*"
58 : tdhock 110
59 :     decomment <- function
60 :     ### Remove comment prefix and join lines of code to form a
61 :     ### documentation string.
62 :     (comments
63 :     ### Character vector of prefixed comment lines.
64 :     ){
65 : tdhock 185 gsub(prefix,"",comments)
66 : tdhock 110 ### String without prefixes or newlines.
67 :     }
68 :    
69 : tdhock 162 forall <- function
70 : tdhock 158 ### For each object in the package that satisfies the criterion
71 :     ### checked by subfun, parse source using FUN and return the resulting
72 : tdhock 87 ### documentation list.
73 :     (FUN,
74 :     ### Function to apply to each element in the package.
75 :     subfun=function(x)TRUE
76 :     ### Function to select subsets of elements of the package, such as
77 :     ### is.function. subfun(x)==TRUE means FUN will be applied to x and
78 :     ### the result will be returned.
79 :     ){
80 : tdhock 138 FUN <- FUN
81 :     f <- function(objs,docs,...){
82 : tdhock 300 if(length(objs)==0)return(list())
83 : tdhock 87 objs <- objs[sapply(objs,subfun)]
84 :     L <- list()
85 : tdhock 162 on.exit(cat(sprintf("Parser Function failed on %s\n",N)))
86 : tdhock 204 for(N in union(names(docs),names(objs))){
87 : tdhock 87 o <- objs[[N]]
88 : tdhock 302 L[[N]] <- FUN(src=getSource(o),
89 : tdhock 89 name=N,objs=objs,o=o,docs=docs,doc=docs[[N]],...)
90 : tdhock 87 }
91 : tdhock 162 on.exit()## remove warning message
92 : tdhock 87 L
93 :     }
94 : tdhock 138 class(f) <- c("allfun","function")
95 :     f
96 : tdhock 87 ### A Parser Function.
97 :     }
98 :    
99 : tdhock 138 ### Print method for functions constructed using forall.
100 :     print.allfun <- function(x,...){
101 :     e <- environment(x)
102 :     cat("Function to apply to every element.\nselector:")
103 :     print(e$subfun)
104 :     cat("processor:")
105 :     print(e$FUN)
106 :     }
107 :    
108 : tdhock 87 ### For each function in the package, do something.
109 :     forfun <- function(FUN)forall(FUN,is.function)
110 :    
111 : tdhock 163 kill.prefix.whitespace <- function
112 :     ### Figure out what the whitespace preceding the example code is, and
113 :     ### then delete that from every line.
114 :     (ex
115 :     ### character vector of example code lines.
116 :     ){
117 :     tlines <- gsub("\\s*","",ex)
118 :     ##tlines <- gsub("#.*","",tlines)
119 :     prefixes <- unique(gsub("\\S.*","",ex[tlines!=""]))
120 :     FIND <- prefixes[which.min(nchar(prefixes))]
121 :     ## Eliminate leading tabulations or 2/4 spaces
122 :     sub(FIND, "", ex)
123 :     ### Character vector of code lines with preceding whitespace removed.
124 :     }
125 :    
126 : tdhock 175 prefixed.lines <- structure(function(src,...){
127 : tdhock 138 ### The primary mechanism of inline documentation is via consecutive
128 :     ### groups of lines matching the specified prefix regular expression
129 :     ### "\code{^### }" (i.e. lines beginning with "\code{### }") are
130 :     ### collected as follows into documentation sections:\describe{
131 :     ### \item{description}{group starting at line 2 in the code}
132 :     ### \item{arguments}{group following each function argument}
133 :     ### \item{value}{group ending at the penultimate line of the code}}
134 :     ### These may be added to by use of the \code{##<<} constructs
135 :     ### described below.
136 :     clines <- grep(prefix,src)
137 :     if(length(clines)==0)return(list())
138 :     bounds <- which(diff(clines)!=1)
139 :     starts <- c(1,bounds+1)
140 :     ends <- c(bounds,length(clines))
141 :     ## detect body of function using paren matching
142 : tdhock 159 code <- gsub("#.*","",src)
143 :     f <- function(ch)cumsum(nchar(gsub(sprintf("[^%s]",ch),"",code)))
144 : tdhock 138 parens <- f("(")-f(")")
145 :     body.begin <- which(diff(parens)<0 & parens[-1]==0)+2
146 : tdhock 175 if(length(body.begin)==0)body.begin <- 1 ## rare cases
147 : tdhock 138 is.arg <- function(){
148 : tdhock 159 gres <- grep("^\\s*#",src[start-1],perl=TRUE)
149 :     0 == length(gres) && start<=body.begin
150 :     }
151 : tdhock 138 res <- list()
152 :     for(i in seq_along(starts)){
153 :     start <- clines[starts[i]]
154 :     end <- clines[ends[i]]
155 : tdhock 304 processed <- gsub("#.*","",gsub("[ }]","",src[(end+1):length(src)]))
156 :     lab <- if(all(processed==""))"value"
157 : tdhock 138 else if(start==2)"description"
158 :     else if(is.arg()){
159 :     ##twutz: strip leading white spaces and brackets and ,
160 : kmpont 305 arg <- gsub("^[ \t(,]*", "", src[start - 1])
161 : tdhock 138 arg <- gsub("^([^=,]*)[=,].*", "\\1", arg)
162 :     ##twutz: remove trailing whitespaces
163 : kmpont 305 arg <- gsub("^([^ \t]*)([ \t]+)$","\\1",arg)
164 : tdhock 138 arg <- gsub("...", "\\dots", arg, fix = TRUE)
165 :     paste("item{",arg,"}",sep="")
166 :     } else {
167 :     next;
168 :     }
169 :     res[[lab]] <- decomment(src[start:end])
170 :     }
171 :     res
172 : tdhock 175 },ex=function(){
173 :     test <- function
174 : tdhock 191 ### the description
175 : tdhock 175 (x,
176 :     ### the first argument
177 :     y ##<< another argument
178 :     ){
179 :     5
180 :     ### the return value
181 :     ##seealso<< foobar
182 : tdhock 138 }
183 : tdhock 302 src <- getSource(test)
184 : tdhock 175 prefixed.lines(src)
185 :     extract.xxx.chunks(src)
186 :     })
187 : tdhock 138
188 : tdhock 154 extract.xxx.chunks <- function # Extract documentation from a function
189 : tdhock 87 ### Given source code of a function, return a list describing inline
190 :     ### documentation in that source code.
191 : tdhock 154 (src,
192 :     ### The source lines of the function to examine, as a character
193 :     ### vector.
194 :     name.fun="(unnamed function)",
195 : tdhock 87 ### The name of the function/chunk to use in warning messages.
196 : tdhock 154 ...
197 :     ### ignored.
198 : tdhock 95 ){
199 : tdhock 87 res <- list()
200 :     ##details<< For simple functions/arguments, the argument may also be
201 :     ## documented by appending \code{##<<} comments on the same line as the
202 :     ## argument name. Mixing this mechanism with \code{###} comment lines for
203 :     ## the same argument is likely to lead to confusion, as the \code{###}
204 :     ## lines are processed first.
205 :     #arg.pat <- paste("^[^=,#]*?([\\w\\.]+)\\s*([=,].*|\\)\\s*)?",
206 :     # "<<\\s*(\\S.*?)\\s*$",
207 :     # sep="##") # paste avoids embedded trigger fooling the system
208 :     #tw: removed first comma
209 :     arg.pat <- paste("^[^=#]*?([\\w\\.]+)\\s*([=,].*|\\)\\s*)?",
210 :     "<<\\s*(\\S.*?)\\s*$",
211 :     sep="##") # paste avoids embedded trigger fooling the system
212 :    
213 :     skeleton.fields <- c("alias","details","keyword","references","author",
214 :     "note","seealso","value","title","description",
215 :     "describe","end")
216 :     ##details<< Additionally, consecutive sections of \code{##} comment
217 :     ## lines beginning with \code{##}\emph{xxx}\code{<<} (where
218 :     ## \emph{xxx} is one of the fields: \code{alias}, \code{details},
219 :     ## \code{keyword}, \code{references}, \code{author}, \code{note},
220 :     ## \code{seealso}, \code{value}, \code{title} or \code{description})
221 :     ## are accumulated and inserted in the relevant part of the .Rd
222 :     ## file.
223 :     ##
224 :     ## For \code{value}, \code{title}, \code{description} and function
225 :     ## arguments, these \emph{append} to any text from "prefix"
226 :     ## (\code{^### }) comment lines, irrespective of the order in the
227 :     ## source.
228 :     ##
229 :     ## When documenting S4 classes, documentation from \code{details}
230 :     ## sections will appear under a section \code{Objects from the Class}. That
231 :     ## section typically includes information about construction methods
232 :     ## as well as other description of class objects (but note that the
233 :     ## class Slots are documented in a separate section).
234 :    
235 :     ## but this should not appear, because separated by a blank line
236 :     extra.regexp <- paste("^\\s*##(",paste(skeleton.fields,collapse="|"),
237 :     ")<<\\s*(.*)$",sep="")
238 :     cont.re <- "^\\s*##\\s*"
239 :     in.describe <- 0
240 :     first.describe <- FALSE
241 :     k <- 1
242 :     in.chunk <- FALSE
243 :     end.chunk <- function(field,payload)
244 :     {
245 :     if ( "alias" == field ){
246 :     ##note<< \code{alias} extras are automatically split at new lines.
247 :     payload <- gsub("\\n+","\\}\n\\\\alias\\{",payload,perl=TRUE)
248 :     chunk.sep <- "}\n\\alias{"
249 :     } else if ( "keyword" == field ){
250 :     ##keyword<< documentation utilities
251 :     ##note<< \code{keyword} extras are automatically split at white space,
252 :     ## as all the valid keywords are single words.
253 :     payload <- gsub("\\s+","\\}\n\\\\keyword\\{",payload,perl=TRUE)
254 :     chunk.sep <- "}\n\\keyword{"
255 :     } else if ( "title" == field ){
256 :     chunk.sep <- " "
257 :     } else if ( "description" == field ){
258 :     chunk.sep <- "\n"
259 :     } else {
260 :     ##details<< Each separate extra section appears as a new
261 :     ## paragraph except that: \itemize{\item empty sections (no
262 :     ## matter how many lines) are ignored;\item \code{alias} and
263 :     ## \code{keyword} sections have special rules;\item
264 :     ## \code{description} should be brief, so all such sections
265 :     ## are concatenated as one paragraph;\item \code{title} should
266 :     ## be one line, so any extra \code{title} sections are
267 :     ## concatenated as a single line with spaces separating the
268 :     ## sections.}
269 :     chunk.sep <- "\n\n"
270 :     }
271 :     chunk.res <- NULL
272 : tdhock 191 if ( !grepl("^\\s*$",payload,perl=TRUE) )
273 : tdhock 87 chunk.res <-
274 :     if ( is.null(res[[field]]) ) payload
275 :     else paste(res[[field]], payload, sep=chunk.sep)
276 :     invisible(chunk.res)
277 :     }
278 : tdhock 154 while ( k <= length(src) ){
279 :     line <- src[k]
280 : tdhock 308 ##print(line)
281 :     ##if(grepl("^$",line))browser()
282 : tdhock 191 if ( grepl(extra.regexp,line,perl=TRUE) ){
283 : tdhock 87 ## we have a new extra chunk - first get field name and any payload
284 :     new.field <- gsub(extra.regexp,"\\1",line,perl=TRUE)
285 :     new.contents <- gsub(extra.regexp,"\\2",line,perl=TRUE)
286 : tdhock 308 ##cat(new.field,"\n-----\n",new.contents,"\n\n")
287 : tdhock 87 ##details<< As a special case, the construct \code{##describe<<} causes
288 :     ## similar processing to the main function arguments to be
289 :     ## applied in order to construct a describe block within the
290 :     ## documentation, for example to describe the members of a
291 :     ## list. All subsequent "same line" \code{##<<} comments go into that
292 :     ## block until terminated by a subsequent \code{##}\emph{xxx}\code{<<} line.
293 :     if ( "describe" == new.field ){
294 :     ##details<< Such regions may be nested, but not in such a way
295 : tdhock 307 ## that the first element in a \code{describe} is another
296 :     ## \code{describe}. Thus there must be at least one
297 :     ## \code{##<<} comment between each pair of
298 :     ## \code{##describe<<} comments.
299 : tdhock 87 if ( first.describe ){
300 :     stop("consecutive ##describe<< at line",k,"in",name.fun)
301 :     } else {
302 :     if ( nzchar(new.contents) ){
303 :     if ( is.null(payload) || 0 == nzchar(payload) ){
304 :     payload <- new.contents
305 :     } else {
306 :     payload <- paste(payload,new.contents,sep="\n\n")
307 :     }
308 :     }
309 :     first.describe <- TRUE
310 :     }
311 :     } else if ( "end" == new.field ){
312 :     ##details<< When nested \code{describe} blocks are used, a comment-only
313 :     ## line with \code{##end<<} terminates the current level only; any
314 :     ## other valid \code{##}\emph{xxx}\code{<<} line terminates
315 :     ## all open describe blocks.
316 :     if ( in.describe>0 ){
317 :     ## terminate current \item and \describe block only
318 :     if ( "value" == cur.field && 1 == in.describe ){
319 :     payload <- paste(payload,"}",sep="")
320 :     } else {
321 :     payload <- paste(payload,"}\n}",sep="")
322 :     }
323 :     in.describe <- in.describe-1;
324 :     } else {
325 :     warning("mismatched ##end<< at line ",k," in ",name.fun)
326 :     }
327 :     if ( nzchar(new.contents) ){
328 :     if ( nzchar(payload) ){
329 :     payload <- paste(payload,new.contents,sep="\n")
330 :     } else {
331 :     payload <- new.contents
332 :     }
333 :     }
334 :     } else {
335 :     ## terminate all open \describe blocks (+1 because of open item)
336 :     if ( 0 < in.describe ){
337 :     if ( "value" != cur.field ){ # value is implicit describe block
338 :     payload <- paste(payload,"}",sep="")
339 :     }
340 :     while ( in.describe>0 ){
341 :     payload <- paste(payload,"}",sep="\n")
342 :     in.describe <- in.describe-1;
343 :     }
344 :     }
345 :     ## finishing any existing payload
346 :     if ( in.chunk ) res[[cur.field]] <- end.chunk(cur.field,payload)
347 :     in.chunk <- TRUE
348 :     cur.field <- new.field
349 :     payload <- new.contents
350 :     ##note<< The "value" section of a .Rd file is implicitly a describe
351 :     ## block and \code{##}\code{value}\code{<<} acts accordingly. Therefore
352 :     ## it automatically enables the describe block itemization (##<< after
353 :     ## list entries).
354 :     if ( "value" == new.field ){
355 :     first.describe <- TRUE;
356 :     }
357 :     }
358 : tdhock 191 } else if ( in.chunk && grepl(cont.re,line,perl=TRUE) ){
359 : tdhock 87 ## append this line to current chunk
360 : tdhock 191 if ( !grepl(prefix,line,perl=TRUE) ){
361 : tdhock 87 ##describe<< Any lines with "\code{### }" at the left hand
362 :     ## margin within the included chunks are handled separately,
363 :     ## so if they appear in the documentation they will appear
364 :     ## before the \code{##}\emph{xxx}\code{<}\code{<} chunks.
365 :     ### This one should not appear.
366 :     stripped <- gsub(cont.re,"",line,perl=TRUE)
367 :     if ( nzchar(payload) ){
368 :     payload <- paste(payload,stripped,sep="\n")
369 :     } else {
370 :     payload <- stripped
371 :     }
372 :     }
373 : tdhock 191 } else if ( grepl(arg.pat,line,perl=TRUE) ){
374 : tdhock 87 not.describe <- (0==in.describe && !first.describe)
375 :     if ( in.chunk && not.describe){
376 :     res[[cur.field]] <- end.chunk(cur.field,payload)
377 :     }
378 :     comment <- gsub(arg.pat,"\\3",line,perl=TRUE);
379 :     arg <- gsub(arg.pat,"\\\\item\\{\\1\\}",line,perl=TRUE)
380 :     in.chunk <- TRUE
381 :     if ( not.describe ){
382 : tdhock 115 ## TDH 2010-06-18 For item{}s in the documentation list names,
383 :     ## we don't need to have a backslash before, so delete it.
384 :     arg <- gsub("^[\\]+","",arg)
385 : tdhock 87 cur.field <- gsub("...","\\dots",arg,fix=TRUE) ##special case for dots
386 :     payload <- comment
387 :     } else {
388 :     ## this is a describe block, so we need to paste with existing
389 :     ## payload as a new \item.
390 :     if ( first.describe ){
391 :     ## for first item, need to add describe block starter
392 :     if ( "value" == cur.field ){
393 :     payload <- paste(payload,"\n",arg,"{",sep="")
394 :     } else {
395 :     payload <- paste(payload,"\\describe{\n",arg,"{",sep="")
396 :     }
397 :     first.describe <- FALSE
398 :     in.describe <- in.describe+1
399 :     } else {
400 :     ## subsequent item - terminate existing and start new
401 :     payload <- paste(payload,"}\n",arg,"{",sep="")
402 :     }
403 :     if ( nzchar(comment) ){
404 :     payload <- paste(payload,comment,sep="")
405 :     }
406 :     }
407 :     } else if ( in.chunk ){
408 :     if ( 0 == in.describe && !first.describe ){
409 :     ## reached an end to current field, but need to wait if in.describe
410 :     res[[cur.field]] <- end.chunk(cur.field,payload)
411 :     in.chunk <- FALSE
412 :     cur.field <- NULL
413 :     payload <- NULL
414 :     }
415 :     }
416 :     k <- k+1
417 :     }
418 :     ## finishing any existing payload
419 :     if ( 0 < in.describe ){
420 :     if ( "value" != cur.field ){ # value is implicit describe block
421 :     payload <- paste(payload,"}",sep="")
422 :     }
423 :     while ( in.describe>0 ){
424 :     payload <- paste(payload,"}",sep="\n")
425 :     in.describe <- in.describe-1;
426 :     }
427 :     }
428 :     if ( in.chunk ) res[[cur.field]] <- end.chunk(cur.field,payload)
429 :     res
430 :     ### Named list of character strings extracted from comments. For each
431 :     ### name N we will look for N\{...\} in the Rd file and replace it
432 :     ### with the string in this list (implemented in modify.Rd.file).
433 :     }
434 :    
435 : kmpont 212 leadingS3generic <- function # check whether function name is an S3 generic
436 :     ### Determines whether a function name looks like an S3 generic function
437 :     (name, ##<< name of function
438 :     env, ##<< environment to search for additional generics
439 :     ...) ##<< ignored here
440 :     {
441 :     ##details<< This function is one of the default parsers, but exposed as
442 :     ## possibly of more general interest. Given a function name of the form
443 :     ## x.y.z it looks for the generic function x applying to objects of class
444 :     ## y.z and also for generic function x.y applying to objects of class z.
445 :     ##
446 :     parts <- strsplit(name, ".", fixed = TRUE)[[1]]
447 :     l <- length(parts)
448 :     if (l > 1) {
449 :     for (i in 1:(l - 1)) {
450 :     ## Look for a generic function (known by the system or defined
451 :     ## in the package) that matches that part of the function name
452 :     generic <- paste(parts[1:i], collapse = ".")
453 :     if (any(generic %in% utils:::getKnownS3generics()) ||
454 :     utils:::findGeneric(generic, env) != "") {
455 : kmpont 305 object <- paste(parts[(i + 1):l], collapse = ".")
456 : kmpont 212 ##details<< Assumes that the first name which matches any known
457 :     ## generics is the target generic function, so if both x and x.y
458 :     ## are generic functions, will assume generic x applying to objects
459 :     ## of class y.z
460 :     ##value<< If a matching generic found returns a list with a single component:
461 :     return(list(.s3method=c(generic, object))) ##<< a character vector containing generic name and object name.
462 :     }
463 :     }
464 :     }
465 :     ##value<< If no matching generic functions are found, returns an empty list.
466 :     list()
467 :     }
468 :    
469 : tdhock 154 ### Parsers for each function that are constructed automatically. This
470 :     ### is a named list, and each element is a parser function for an
471 :     ### individual object.
472 :     forfun.parsers <-
473 :     list(prefixed.lines=prefixed.lines,
474 :     extract.xxx.chunks=extract.xxx.chunks,
475 :     ## title from first line of function def
476 : tdhock 191 title.from.firstline=function(src,...){
477 : tdhock 195 first <- src[1]
478 :     if(!is.character(first))return(list())
479 :     if(!grepl("#",first))return(list())
480 :     list(title=gsub("[^#]*#\\s*(.*)","\\1",first,perl=TRUE))
481 : tdhock 154 },
482 :     ## PhG: it is tests/FUN.R!!! I would like more flexibility here
483 :     ## please, let me choose which dir to use for examples!
484 :     ## Get examples for FUN from the file tests/FUN.R
485 :     examples.from.testfile=function(name,...){
486 :     tsubdir <- getOption("inlinedocs.exdir")
487 :     if (is.null(tsubdir)) tsubdir <- "tests" # Default value
488 :     tfile <- file.path("..",tsubdir,paste(name,".R",sep=""))
489 :     if(file.exists(tfile))
490 : tdhock 185 list(examples=readLines(tfile))
491 : tdhock 154 else list()
492 :     },
493 :     definition.from.source=function(doc,src,...){
494 :     def <- doc$definition
495 :     is.empty <- function(x)is.null(x)||x==""
496 :     if(is.empty(def) && !is.empty(src))
497 : tdhock 185 list(definition=src)
498 : tdhock 154 else list()
499 :     })
500 :    
501 :     ### List of Parser Functions that can be applied to any object.
502 :     forall.parsers <-
503 :     list(## Fill in author from DESCRIPTION and titles.
504 :     author.from.description=function(desc,...){
505 : tdhock 308 list(author=desc[,"Author"])
506 : tdhock 154 },
507 :     ## The format section sometimes causes problems, so erase it.
508 :     erase.format=function(...){
509 :     list(format="")
510 :     },
511 :     ## Convert the function name to a title.
512 :     title.from.name=function(name,doc,...){
513 :     if("title"%in%names(doc))list() else
514 :     list(title=gsub("[._]"," ",name))
515 :     },
516 :     ## PhG: here is what I propose for examples code in the 'ex' attribute
517 :     examples.in.attr = function (name, o, ...) {
518 :     ex <- attr(o, "ex")
519 :     if (!is.null(ex)) {
520 :     ## Special case for code contained in a function
521 :     if (inherits(ex, "function")) {
522 :     ## If source is available, start from there
523 : tdhock 302 src <- getSource(ex)
524 : tdhock 154 if (!is.null(src)) {
525 :     ex <- src
526 :     } else { ## Use the body of the function
527 :     ex <- deparse(body(ex))
528 :     }
529 :     ## Eliminate leading and trailing code
530 :     ex <- ex[-c(1, length(ex))]
531 : tdhock 303 if(ex[1]=="{")ex <- ex[-1]
532 : tdhock 163 ## all the prefixes
533 :     ex <- kill.prefix.whitespace(ex)
534 : tdhock 154 ## Add an empty line before and after example
535 :     ex <- c("", ex, "")
536 :     }
537 : tdhock 185 list(examples = ex)
538 : tdhock 154 } else list()
539 : tdhock 200 },collapse=function(doc,...){
540 :     L <- lapply(doc,paste,collapse="\n")
541 :     L$.overwrite <- TRUE
542 :     L
543 : kmpont 212 },tag.s3methods=leadingS3generic
544 : tdhock 196 )
545 : tdhock 154
546 :     ### List of parser functions that operate on single objects. This list
547 :     ### is useful for testing these functions.
548 : tdhock 196 lonely <- structure(c(forall.parsers,forfun.parsers),ex=function(){
549 : tdhock 154 f <- function # title
550 :     ### description
551 :     (x, ##<< arg x
552 :     y
553 :     ### arg y
554 :     ){
555 :     ##value<< a list with elements
556 :     list(x=x, ##<< original x value
557 :     y=y, ##<< original y value
558 :     sum=x+y) ##<< their sum
559 :     ##end<<
560 :     }
561 : tdhock 302 src <- getSource(f)
562 : tdhock 154 lonely$extract.xxx.chunks(src)
563 :     lonely$prefixed.lines(src)
564 : tdhock 196 })
565 : tdhock 154
566 :     extra.code.docs <- function # Extract documentation from code chunks
567 :     ### Parse R code to extract inline documentation from comments around
568 :     ### each function. These are not able to be retreived simply by
569 :     ### looking at the "source" attribute. This is a Parser Function that
570 :     ### can be used in the parser list of package.skeleton.dx(). TODO:
571 :     ### Modularize this into separate Parsers Functions for S4 classes,
572 :     ### prefixes, ##<<blocks, etc. Right now it is not very clean!
573 :     (code,
574 :     ### Code lines in a character vector containing multiple R objects to
575 :     ### parse for documentation.
576 :     objs,
577 :     ### The objects defined in the code.
578 :     ...
579 :     ### ignored
580 :     ){
581 :     parsed <- extract.file.parse(code)
582 :     extract.docs.try <- function(o,on)
583 :     {
584 :     ## Note: we could use parsed information here too, but that
585 :     ## would produce different results for setMethodS3 etc.
586 :     doc <- list()
587 :     if ( !is.null(parsed[[on]]) ){
588 :     if ( !is.na(parsed[[on]]@code[1]) ){ # no code given for generics
589 : tdhock 185 doc$definition <- paste(parsed[[on]]@code)
590 : tdhock 154 }
591 :     if(!"description"%in%names(doc) && !is.na(parsed[[on]]@description) ){
592 :     doc$description <- parsed[[on]]@description
593 :     }
594 : kmpont 212 ## if ( "setMethodS3" == parsed[[on]]@created ){
595 :     ## gen <- leadingS3generic(on,topenv())
596 :     ## if ( 0 < length(gen) ){
597 :     ## doc$.s3method <- gen$.s3method
598 :     ## cat("S3method(",gen$.s3method[1],",",gen$.s3method[2],")\n",sep="")
599 :     ## }
600 :     ## }
601 : tdhock 154 }
602 :     if("title" %in% names(doc) && !"description" %in% names(doc) ){
603 :     ## For short functions having both would duplicate, but a
604 :     ## description is required. Therefore automatically copy title
605 :     ## across to avoid errors at package build time.
606 :     doc$description <- doc$title
607 :     }
608 :     doc
609 :     }
610 :     extract.docs <- function(on){
611 :     res <- try({o <- objs[[on]]
612 :     extract.docs.try(o, on)},FALSE)
613 :     if(class(res)=="try-error"){
614 :     cat("Failed to extract docs for: ",on,"\n\n")
615 :     list()
616 :     } else if(0 == length(res) && inherits(objs[[on]],"standardGeneric")){
617 :     NULL
618 :     } else if(0 == length(res) && "function" %in% class(o)
619 : tdhock 302 && 1 == length(osource <- getSource(o))
620 : tdhock 191 && grepl(paste("UseMethod(",on,")",sep="\""),osource)
621 : tdhock 154 ){
622 :     ## phew - this should only pick up R.oo S3 generic definitions like:
623 :     ## attr(*, "source")= chr "function(...) UseMethod(\"select\")"
624 :     NULL
625 :     } else res
626 :     }
627 :     doc.names <- names(objs)
628 :     res <- sapply(doc.names,extract.docs,simplify=FALSE)
629 :     ## Special processing for S4 classes as they do not appear in normal ls()
630 :     for ( nn in names(parsed) ){
631 :     if ( parsed[[nn]]@created == "setClass" ){
632 :     S4class.docs <- extract.docs.setClass(parsed[[nn]])
633 :     docname <- paste(nn,"class",sep="-")
634 :     if ( is.null(res[[docname]]) ){
635 :     res[[docname]] <- S4class.docs
636 :     doc.names <- c(doc.names,docname)
637 :     } else {
638 :     stop(nn," appears as both S4 class and some other definition")
639 :     }
640 :     }
641 :     }
642 :     inherit.docs <- function(on){
643 :     in.res <- res[[on]]
644 :     if ( !is.null(parsed[[on]]) ){
645 :     for ( parent in parsed[[on]]@parent ){
646 :     if ( !is.na(parent) ){
647 :     if ( is.null(in.res) ){
648 :     in.res <- res[[parent]]
649 :     } else if ( parent %in% names(res) ){
650 :     parent.docs <- res[[parent]]
651 :     for ( nn in names(parent.docs) ){
652 :     if ( !nn %in% names(in.res) ){
653 :     in.res[[nn]] <- parent.docs[[nn]]
654 :     }
655 :     }
656 :     }
657 :     }
658 :     }
659 :     }
660 :     invisible(in.res)
661 :     }
662 :     all.done <- FALSE
663 :     while ( !all.done ){
664 :     res1 <- sapply(doc.names,inherit.docs,simplify=FALSE)
665 :     all.done <- identical(res1,res)
666 :     res <- res1
667 :     }
668 :     ## now strip out any generics (which have value NULL in res):
669 :     res.not.null <- sapply(res,function(x){!is.null(x)})
670 :     if ( 0 < length(res.not.null) && length(res.not.null) < length(res) ){
671 :     res <- res[res.not.null]
672 :     }
673 :     res
674 :     ### named list of lists, one for each object to document.
675 :     }
676 :    
677 :     ### List of parsers to use by default with package.skeleton.dx.
678 :     default.parsers <-
679 :     c(extra.code.docs=extra.code.docs, ## TODO: cleanup!
680 :     sapply(forfun.parsers,forfun),
681 :     edit.package.file=function(desc,...){
682 :     in.details <- setdiff(colnames(desc),"Description")
683 : tdhock 185 details <- sprintf("%s: \\tab %s\\cr",in.details,desc[,in.details])
684 : tdhock 154 L <-
685 :     list(list(title=desc[,"Title"],
686 :     description=desc[,"Description"],
687 : tdhock 195 `tabular{ll}`=details))
688 : tdhock 154 names(L) <- paste(desc[,"Package"],"-package",sep="")
689 :     L
690 : tdhock 197 },
691 :     sapply(forall.parsers,forall)
692 : tdhock 186 )
693 : tdhock 154
694 : tdhock 87 setClass("DocLink", # Link documentation among related functions
695 :     ### The \code{.DocLink} class provides the basis for hooking together
696 :     ### documentation of related classes/functions/objects. The aim is that
697 : kmpont 212 ### documentation sections missing from the child are inherited from
698 :     ### the parent class.
699 : tdhock 87 representation(name="character", ##<< name of object
700 :     created="character", ##<< how created
701 :     parent="character", ##<< parent class or NA
702 :     code="character", ##<< actual source lines
703 :     description="character") ##<< preceding description block
704 :     )
705 :    
706 :     extract.file.parse <- function # File content analysis
707 :     ### Using the base \code{\link{parse}} function, analyse the file to link
708 :     ### preceding "prefix" comments to each active chunk. Those comments form
709 :     ### the default description for that chunk. The analysis also looks for
710 :     ### S4 class "setClass" calls and R.oo setConstructorS3 and setMethodS3
711 :     ### calls in order to link the documentation of those properly.
712 :     (code
713 :     ### Lines of R source code in a character vector - note that any
714 :     ### nested \code{source} statements are \emph{ignored} when scanning
715 :     ### for class definitions.
716 :     ){
717 :     res <- list()
718 :     old.opt <- options(keep.source=TRUE)
719 :     parsed <- try(parse(text=code))
720 :     options(old.opt)
721 :     if ( inherits(parsed,"try-error") ){
722 :     stop("parse failed with error:\n",parsed)
723 :     }
724 :     chunks <- attr(parsed,"srcref")
725 :     last.end <- 0
726 :     for ( k in 1:length(parsed) ){
727 :     start <- chunks[[k]][1]
728 :     ##details<< If the definition chunk does not contain a
729 :     ## description, any immediately preceding sequence consecutive
730 :     ## "prefix" lines will be used instead.
731 :     default.description <- NULL
732 :     while ( start > last.end+1
733 : tdhock 191 && grepl(prefix,code[start-1],perl=TRUE) ){
734 : tdhock 87 start <- start-1
735 :     }
736 :     if ( start < chunks[[k]][1] ){
737 :     default.description <- decomment(code[start:(chunks[[k]][1]-1)])
738 :     } else {
739 :     default.description <- NA_character_;
740 :     }
741 :     ##details<< Class and method definitions can take several forms,
742 :     ## determined by expression type: \describe{
743 :     ## \item{assignment (<-)}{Ordinary assignment of value/function;}
744 :     ## \item{setClass}{Definition of S4 class;}
745 :     ## \item{setConstructorS3}{Definition of S3 class using R.oo package;}
746 :     ## \item{setMethodS3}{Definition of method for S3 class using R.oo package.}}
747 :     ## Additionally, the value may be a name of a function defined elsewhere,
748 :     ## in which case the documentation should be copied from that other definition.
749 :     ## This is handled using the concept of documentation links.
750 :     lang <- parsed[[k]]
751 :     chars <- as.character(lang)
752 :     expr.type <- chars[1]
753 :     parent <- NA_character_
754 :    
755 :     if ( expr.type == "<-" || expr.type == "setConstructorS3" || expr.type == "setClass" ){
756 :     object.name <- chars[2]
757 :     ## If the function definition is not embedded within the call, then
758 :     ## the parent is that function. Test whether the the third value
759 :     ## looks like a name and add it to parents if so.
760 : tdhock 191 if ( grepl("^[\\._\\w]+$",chars[3],perl=TRUE) ){
761 : tdhock 87 parent <- chars[3]
762 :     }
763 :     res[[object.name]] <- new("DocLink",name=object.name,
764 :     created=expr.type,
765 :     parent=parent,
766 :     code=paste(chunks[[k]],sep=""),
767 :     description=default.description)
768 :     } else if ( expr.type == "setMethodS3" ){
769 :     ##details<< The \code{setMethodS3} calls introduce additional
770 :     ## complexity: they will define an additional S3 generic (which
771 :     ## needs documentation to avoid warnings at package build time)
772 :     ## unless one already exists. This also is handled by "linking"
773 :     ## documentation. A previously unseen generic is linked to the
774 :     ## first defining instances, subsequent definitions of that generic
775 :     ## also link back to the first defining instance.
776 :     generic.name <- chars[2]
777 :     object.name <- paste(generic.name,chars[3],sep=".")
778 :     if ( is.null(res[[generic.name]]) ){
779 :     generic.desc <- paste("Generic method behind \\code{\\link{",object.name,"}}",sep="")
780 :     res[[generic.name]] <- new("DocLink",
781 :     name=generic.name,
782 :     created=expr.type,
783 :     parent=object.name,
784 :     code=NA_character_,
785 :     description=generic.desc)
786 :     } else {
787 :     parent <- res[[generic.name]]@parent
788 :     }
789 :     ## If the function definition is not embedded within the call, then
790 :     ## the parent is that function. Test whether the the fourth value
791 :     ## looks like a name and add it to parents if so.
792 : tdhock 191 if ( grepl("^[\\._\\w]+$",chars[4],perl=TRUE) ){
793 : tdhock 87 parent <- c(chars[4],parent)
794 :     }
795 :     res[[object.name]] <- new("DocLink",name=object.name,
796 :     created=expr.type,
797 :     parent=parent,
798 :     code=paste(chunks[[k]],sep=""),
799 :     description=default.description)
800 :     } else {
801 :     ## Not sure what to do with these yet. Need to deal with setMethod, setAs etc.
802 :     }
803 :     }
804 :     invisible(res)
805 :     ### Returns an invisible list of .DocLink objects.
806 :     }
807 :    
808 :     extract.docs.setClass <- function # S4 class inline documentation
809 :     ### Using the same conventions as for functions, definitions of S4 classes
810 :     ### in the form \code{setClass("classname",\dots)} are also located and
811 :     ### scanned for inline comments.
812 : tdhock 93 (doc.link
813 : tdhock 87 ### DocLink object as created by \code{\link{extract.file.parse}}.
814 :     ### Note that \code{source} statements are \emph{ignored} when scanning for
815 :     ### class definitions.
816 : tdhock 93 ){
817 : tdhock 87 chunk.source <- doc.link@code
818 :     ##details<<
819 :     ## Extraction of S4 class documentation is currently limited to expressions
820 :     ## within the source code which have first line starting with
821 :     ## \code{setClass("classname"}. These are located from the source file
822 :     ## (allowing also for white space around the \code{setClass} and \code{(}).
823 :     ## Note that \code{"classname"} must be a quoted character string;
824 :     ## expressions returning such a string are not matched.
825 :     class.name <- doc.link@name
826 :    
827 :     ##details<< For class definitions, the slots (elements of the
828 :     ## \code{representation} list) fill the role of function
829 :     ## arguments, so may be documented by \code{##<<} comments on
830 :     ## the same line or \code{### } comments at the beginning of the
831 :     ## following line.
832 :     f.n <- paste(class.name,"class",sep="-")
833 : tdhock 154 docs <- extract.xxx.chunks(chunk.source,f.n)
834 : tdhock 138 ## also apply source parsing functions that I separated out into
835 :     ## separate functions
836 :     docs <- combine(docs,lonely$prefixed.lines(chunk.source))
837 :     docs$title <- lonely$title.from.firstline(chunk.source)
838 : tdhock 87 ##details<<
839 : kmpont 305 ## If there is no explicit title on the first line of setClass, then
840 :     ## one is made up from the class name.
841 :     if ( 0 == length(docs$title) ){
842 :     docs$title <- list(title=paste(class.name,"S4 class"))
843 :     }
844 :     ##details<<
845 : tdhock 87 ## The class definition skeleton includes an \code{Objects from the Class}
846 :     ## section, to which any \code{##details<<} documentation chunks are
847 :     ## written. It is given a vanilla content if there are no specific
848 :     ## \code{##details<<} documentation chunks.
849 :     if ( is.null(docs[["details"]]) ){
850 :     docs[["details"]] <-
851 :     paste("Objects can be created by calls of the form \\code{new(",
852 :     class.name," ...)}",sep="")
853 :     }
854 :     docs[["section{Objects from the Class}"]] <- docs[["details"]]
855 :     ## seealso has a skeleton line not marked by ~ .. ~, so have to suppress
856 :     if ( is.null(docs[["seealso"]]) ){
857 :     docs[["seealso"]] <- ""
858 :     }
859 :     if ( is.null(docs[["alias"]]) ){
860 :     docs[["alias"]] <- class.name
861 :     }
862 :     if ( is.null(docs[["description"]]) ){
863 :     docs[["description"]] <- doc.link@description
864 :     }
865 :     invisible(docs)
866 :     }
867 : tdhock 105
868 : tdhock 109 apply.parsers <- function
869 :     ### Parse code to r objs, then run all the parsers and return the
870 :     ### documentation list.
871 : tdhock 105 (code,
872 :     ### Character vector of code lines.
873 : tdhock 109 parsers=default.parsers,
874 : tdhock 105 ### List of Parser Functions.
875 : tdhock 109 verbose=FALSE,
876 : tdhock 105 ### Echo names of Parser Functions?
877 :     ...
878 :     ### Additional arguments to pass to Parser Functions.
879 :     ){
880 :     e <- new.env()
881 : kmpont 212 ## KMP 2011-03-09 fix problem with DocLink when inlinedocs ran on itself
882 :     ## Error in assignClassDef(Class, classDef, where) :
883 :     ## Class "DocLink" has a locked definition in package "inlinedocs"
884 :     ## Traced to "where" argument in setClassDef which defaults to topenv()
885 :     ## which in turn is inlinedocs when processing inlinedocs package, hence
886 :     ## the clash. The following works (under R 2.12.2), so that the topenv()
887 :     ## now finds e before finding the inlinedocs environment.
888 : kmpont 213 old <- options(keep.source=TRUE,topLevelEnvironment=e)
889 : tdhock 109 on.exit(options(old))
890 : tdhock 110 exprs <- parse(text=code)
891 : tdhock 248 ## TDH 2011-04-07 set this so that no warnings about creating a fake
892 :     ## package when we try to process S4 classes defined in code
893 :     e$.packageName <- "inlinedocs.processor"
894 : tdhock 152 for (i in exprs){
895 : tdhock 248 eval(i, e)
896 : tdhock 152 }
897 : tdhock 105 objs <- sapply(ls(e),get,e,simplify=FALSE)
898 :    
899 :     docs <- list()
900 : kmpont 305
901 : tdhock 109 ## apply parsers in sequence to code and objs
902 : tdhock 312 if(verbose)cat("Applying parsers:\n")
903 : tdhock 105 for(i in seq_along(parsers)){
904 :     N <- names(parsers[i])
905 :     if(verbose){
906 :     if(is.character(N) && N!=""){
907 : tdhock 312 cat(N,"\n",sep="")
908 :     }else cat('.\n')
909 : tdhock 105 }
910 :     p <- parsers[[i]]
911 :     ## This is the argument list that each parser receives:
912 : tdhock 200 L <- p(code=code,objs=objs,docs=docs,env=e,...)
913 : markus 324 #print(paste(L,"\n"))
914 : tdhock 335 #if(N=="exclude")browser()
915 : tdhock 105 docs <- combine(docs,L)
916 :     }
917 : tdhock 196 ## post-process to collapse all character vectors
918 :     for(i in seq_along(docs)){
919 :     for(j in seq_along(docs[[i]])){
920 : tdhock 198 if(names(docs[[i]])[j]!=".s3method")
921 : tdhock 196 docs[[i]][[j]] <- paste(docs[[i]][[j]],collapse="\n")
922 :     }
923 : markus 324 }
924 : tdhock 105 if(verbose)cat("\n")
925 : markus 324 return(docs)
926 : tdhock 105 ### A list of extracted documentation from code.
927 :     }
928 : tdhock 109
929 : tdhock 138 ### Names of Parser Functions that operate on the desc arg.
930 :     descfile.names <- c("author.from.description","edit.package.file")
931 :    
932 :     ### Names of Parser Functions that do NOT use the desc arg.
933 :     non.descfile.names <-
934 :     names(default.parsers)[!names(default.parsers)%in%descfile.names]
935 :    
936 : tdhock 118 ### Parsers that operate only on R code, independently of the
937 :     ### description file.
938 : tdhock 138 nondesc.parsers <- default.parsers[non.descfile.names]
939 : tdhock 118
940 : tdhock 209 extract.docs.file <- structure(function
941 : tdhock 109 ### Apply all parsers relevant to extract info from just 1 code file.
942 :     (f,
943 :     ### File name of R code to read and parse.
944 : tdhock 203 parsers=NULL,
945 : tdhock 109 ### Parser Functions to use to parse the code and extract
946 :     ### documentation.
947 :     ...
948 :     ### Other arguments to pass to Parser Functions.
949 :     ){
950 : tdhock 203 if(is.null(parsers))parsers <- nondesc.parsers
951 : tdhock 109 apply.parsers(readLines(f),parsers,verbose=FALSE,...)
952 : tdhock 209 },ex=function(){
953 :     f <- system.file("silly","R","silly.R",package="inlinedocs")
954 : tdhock 248 extract.docs.file(f)
955 : tdhock 209 })
956 : tdhock 248

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