reprompt: Update the documentation of a topic

Description

Examine the documentation of functions, methods or classes and update it with additional arguments, aliases, methods or slots, as appropriate. This is an extention of the promptXXX() family of functions.

Usage

reprompt(object, infile = NULL, Rdtext = NULL, final = TRUE, 
         type = NULL, package = NULL, methods = NULL, verbose = TRUE, 
         filename = NULL, sec_copy = TRUE, edit = FALSE, ...)

Value

if filename is a character string or NULL, the name of the file to which the updated shell was written.

Otherwise, the Rd text is returned as a character vector.

Arguments

object: the object whose documentation is to be updated, such as a string, a function, a help topic, a parsed Rd object, see `Details'.
infile: a file name containing Rd documentation, see `Details'.
Rdtext: a character string with Rd formatted text, see `Details'.
final: logical, if TRUE modifies the output of prompt so that the package can be built.
type: type of topic, such as "methods" or "class", see `Details'.
package: package name; document only objects defined in this package, especially useful for methods, see `Details'.
methods: used for documentation of S4 methods only, rarely needed even for them. This argument is passed on to promptMethods, see its documentation for details.
verbose: if TRUE print messages on the screen.
filename: name of the file for the generated Rd content; if NULL, a suitable file name is generated, if TRUE it will be set to infile, if FALSE the Rd text is returned, see `Details'.
...: currently not used.
sec_copy: if TRUE copy Rd contents of unchanged sections in the result, see Details.
edit: if TRUE call file.edit just before returning. This argument is ignored if filename is FALSE.

Author

Georgi N. Boshnakov

Details

The typical use of reprompt is with one argument, as in


    reprompt(infile = "./Rdpack/man/reprompt.Rd")
    reprompt(reprompt)
    reprompt("reprompt")

reprompt updates the requested documentation and writes the new Rd file in the current working directory. When argument infile is used, the descriptions of all objects described in the input file are updated. When an object or its name is given, reprompt looks first for installed documentation and processes it in the same way as in the case of infile. If there is no documentation for the object, reprompt creates a skeleton Rd file similar to the one produced by the base R functions from the prompt family (if final = TRUE, the default, it modifies the output of prompt(), so that the package can be built).

To document a function, say myfun, in an existing Rd file, just add myfun() to the usage section in the file and call reprompt() on that file. Put quotes around the function name if it is non-syntactic. For replacement functions (functions with names ending in <-) reprompt() will insert the proper usage statement. For example, if the signature of xxx<- is (x, ..., value) then both, "xxx<-"() and xxx() <- value will be replaced by xxx(x, ...) <- value.

For S4 methods and classes the argument "package" is often needed to restrict the output to methods defined in the package of interest.


    reprompt("myfun-methods")
    reprompt("[<--methods", package = "mypackage")             
    reprompt("[<-", type = "methods", package = "mypackage") # same
    reprompt("myclass", type = "class", package = "mypackage") 
    reprompt("myclass-class", package = "mypackage")         # same

Without the "package" argument the reprompt for "[<-" would give all methods defined by loaded packages at the time of the call.

Currently reprompt functionality is not implemented for topic "package" but if object has the form "name-package" (or the equivalent with argument topic) and there is no documentation for package?name, reprompt calls promptPackageSexpr to create the required shell. Note that the shell produced by promptPackageSexpr does not need `reprompting' since the automatically generated information is included by \Sexpr's, not literal strings.

Below are the details.

Typically, only one of object, infile, and Rdtext is supplied. Warning messages are issued if this is not the case.

The object must have been made available by the time when reprompt() is issued. If the object is in a package this is typically achieved by a library() command.

object may be a function or a name, as accepted by the ? operator. If it has the form "name-class" and "name-methods" a documentation shell for class "name" or the methods for generic function "name" will be examined/created. Alternatively, argument type may be set to "class" or "methods" to achieve the same effect.

infile specifies a documentation file to be updated. If it contains the documentation for one or more functions, reprompt examines their usage statements and updates them if they have changed. It also adds arguments to the "arguments" section if not all arguments in the usage statements have entries there. If infile describes the methods of a function or a class, the checks made are as appropriate for them. For example, new methods and/or slots are added to the corresponding sections.

It is all too easy in interactive use to forget to name the infile argument, compare
reprompt("./man/reprompt.Rd") vs. reprompt(infile = "./man/reprompt.Rd")).
A convenience feature is that if infile is missing and object is a character string ending in ".Rd" and containing a forward slash (i.e. it looks like Rd file in a directory), then it is taken to be infile.

Rdtext is similar to infile but the Rd content is taken from a character vector.

If Rd content is supplied by infile or Rdtext, reprompt uses it as a basis for comparison. Otherwise it tries to find installed documentation for the object or topic requested. If that fails, it resorts to one of the promptXXX functions to generate a documentation shell. If that also fails, the requested object or topic does not exist.

If the above succeeds, the function examines the current definition of the requested object(s), methods, or class and amends the documentation with any additional items it finds.

For Rd objects describing one or more functions, the usage expressions are checked and replaced, if they have changed. Arguments appearing in one or more usage expressions and missing from section "Arguments" are amended to it with content "Describe ..." and a message is printed. Arguments no longer in the usage statements are NOT removed but a message is issued to alert the user. Alias sections are inserted for any functions with "usage" but without "alias" section.

If filename is a character string, it is used as is, so any path should be included in the string. Argument filename usuallly is omitted since the automatically generated file name is suitable in most cases. Exceptions are functions with non-standard names (such as replacement functions whose names end in "<-") for which the generated file names may not be acceptable on some operating systems.

If filename is missing or NULL, a suitable name is generated as follows. If infile is supplied, filename is set to a file with the same name in the current working directory (i.e. any path information in infile is dropped). Otherwise, filename is obtained by appending the name tag of the Rd object with ".Rd".

If filename is TRUE, it is set to infile or, if infile is missing or NULL, a suitable name is generated as above. This can be used to change infile in place.

If filename is FALSE, the Rd text is returned as a character vector and not written to a file.

If edit is TRUE, the reprompted file is opened in an editor, see also ereprompt (`e' for `edit') which is like reprompt but has as default edit = TRUE and some other related settings.

file.edit() is used to call the editor. Which editor is opened depends on the OS and on the user configuration. RStudio users will probably prefer the 'Reprompt' add-in or the underlying function RStudio_reprompt. Emacs users would normally have set up emacsclient as their editor and this is automatically done by EMACS/ESS (even on Windows).

If argument sec_copy is TRUE (the default), reprompt will, effectively, copy the contents of (some) unchanged sections, thus ensuring that they are exactly the same as in the original. This needs additional work, since parsing an Rd file and then exporting the created Rd object to an Rd file does not necessarilly produce an identical Rd file (some escape sequences may be changed in the process, for example). Even though the new version should be functionally equivalent to the original, such changes are usually not desirable. For example, if such changes creep into the Details section (which reprompt never changes) the user may be annoyed or worried.

Examples

Run this code

## note: usage of reprompt() is simple.  the examples below are bulky
##       because they simulate various usage scenarios with commands,
##       while in normal usage they would be due to editing.

## change to a temporary directory to avoid clogging up user's
cur_wd <- getwd()
tmpdir <- tempdir()
setwd(tmpdir)

## as for prompt() the default is to save in current dir as "seq.Rd".
## the argument here is a function, reprompt finds its doc and
## updates all objects described along with `seq'.
## (In this case there is nothing to update, we have not changed `seq'.)

fnseq <- reprompt(seq)

## let's parse the saved Rd file (the filename is returned by reprompt)
rdoseq <- tools::parse_Rd(fnseq)   # parse fnseq to see the result.
Rdo_show(rdoseq)

## we replace usage statements with wrong ones for illustration.
## (note there is an S3 method along with the functions)
dummy_usage <- char2Rdpiece(paste("seq()", "\\method{seq}{default}()",
                   "seq.int()", "seq_along()", "seq_len()", sep="\n"),
                   "usage")
rdoseq_dummy <- Rdo_replace_section(rdoseq, dummy_usage)
Rdo_show(rdoseq_dummy)  # usage statements are wrong

reprompt(rdoseq_dummy, file = "seqA.Rd")
Rdo_show(tools::parse_Rd("seqA.Rd"))  # usage ok after reprompt

## define function myseq() 
myseq <- function(from, to, x){
    if(to < 0) {
        seq(from = from, to = length(x) + to)
    } else seq(from, to)
}

## we wish to describe  myseq() along with seq();
##    it is sufficient to put myseq() in the usage section
##    and let reprompt() do the rest
rdo2 <- Rdo_modify_simple(rdoseq, "myseq()", "usage")
reprompt(rdo2, file = "seqB.Rd")  # updates usage of myseq

## show the rendered result:
Rdo_show(tools::parse_Rd("seqB.Rd"))

## Run this if you wish to see the Rd file:
##   file.show("seqB.Rd")

reprompt(infile = "seq.Rd", filename = "seq2.Rd")
reprompt(infile = "seq2.Rd", filename = "seq3.Rd")

## Rd objects for installed help may need some tidying for human editing.
#hseq_inst <- help("seq")
#rdo <- utils:::.getHelpFile(hseq_inst)
rdo <- Rdo_fetch("seq", "base")
rdo
rdo <- Rdpack:::.Rd_tidy(rdo)          # tidy up (e.g. insert new lines
                                       #          for human readers)
reprompt(rdo) # rdo and rdoseq are equivalent
all.equal(reprompt(rdo), reprompt(rdoseq)) # TRUE

## clean up 
unlink("seq.Rd")         # remove temporary files
unlink("seq2.Rd")
unlink("seq3.Rd")
unlink("seqA.Rd")
unlink("seqB.Rd")

setwd(cur_wd)            # restore user's working directory
unlink(tmpdir)

Run the code above in your browser using DataLab