== TODO

Rd generation for "[.class" as in quadmod.

Attempt doc list to Rd list conversion, so we can generate Rd using
tools:::print.Rd. See code in etc/parseRd.R

Try to get more tests working: R CMD check and Rd conversion to
doclist checks.

Rd generation for attrSetter<- functions?

Functions with {} as default arguments => error. See
etc/default-function.R

== 1.9.2 --- 28 Jan 2013

Pierre Neuvial reports that concatenating files with comments at the
last line sometimes results in unexpected documentation associated
with the first object defined in the next file. Thus now we issue a
warning whenever there is a comment on the last line of an .R file.

check for existence of files in removeAliasesfrom.Rd.file, since they
sometimes will not exist due to do.not.generate.

== 1.9.1 --- 21 May 2012

use --as-cran for tests!

R CMD check --as-cran inlinedocs_1.9.tar.gz had 
* checking Rd cross-references ... WARNING
... this seems to be related to \link{} tags in Rd?
... or some strange bug in R CMD check?
https://bugs.r-project.org/bugzilla3/show_bug.cgi?id=14875

pause for 4 seconds in example for do.not.generate. On windows the
filesystem records mtimes with a resolution of 2 seconds, so if we
want to see that a file has changed we need to wait at least this
long.

== 1.9 --- 13 Jan 2012 -> 5 April 2012

move code for generating a DESCRIPTION file into package.skeleton.dx.

doc bugfix for some \code{} chunks in inlinedocs that were
disappearing.

Added inst/silly/NAMESPACE to avoid warnings on package check, and
copy this over to all the mini test packages that are created for each
of the inst/testfiles.

Use \s space instead of \W in regexp for "delete empty sections to
suppress warnings in R CMD check." This was causing a bug for
i.e. \code{+} and \code{%*%} in the doc for the FUN argument of the
apply function.

do.not.generate("rd1","rd2") returns a Parser Function which will
cause rd1.Rd and rd2.Rd to not be generated, so you can write these Rd
files by hand.

defaults for DESCRIPTION entries.

Changes from Markus Muller, excludePattern argument to package.skeleton.dx

Suggestion from Mark Thyer to not put a return after EVERY comma in
the usage section. Instead, we parse the expression and use format to
add line breaks.

== 1.8, 21 Oct 2011

Remove inlinedocExample<- since this was deprecated and now giving a
WARNING on package check.

Take \author for Rd files from Author: section of DESCRIPTION, instead
of Maintainer: section.

bugfix for getSource, for R CMD check with R.2.14.

== 1.7

blank lines allowed after value comments.

usage section broken over lines in modify.Rd.file for prettying pdf
output.

bugfix for when there are no objects defined in the
package. previously forall functions would stop with an error, now we
test for this condition before the line where we would get the error.

getSource(FUN) function from Duncan Murdoch used to extract function
source code, instead of attr(FUN,"source") for compatibility with
R>=2.14.

== 1.6

Only do R CMD check in tests when we run interactively. If not
CRAN/R-Forge gives a weird error.

Remove support for examples.after.return.

Multiparagraph descriptions and unit test.

Description before unit test.

Suppress warnings when we attempt to detach loaded packages, in order
to avoid these warnings:

> warnings()
Warning messages:
1: In FUN(X[[2L]], ...) :
  Created a package name, "2011-05-11 14:11:30", when none found
2: In FUN(X[[2L]], ...) :
  Created a package name, "2011-05-11 14:11:30", when none found
...

Fix warning when S4 classes are defined in package code, by setting
the .packageName variable in the env in apply.parsers, so we don't get
this warning anymore

Warning in getPackageName(where) :
  Created a package name, "2011-04-07 15:29:30", when none found

== 1.5.1

Bugfix for S4 classes by setting the globalenv in apply.parsers to the
env where we eval the pkg code. We were getting this error

Error in assign(mname, def, where) : 
  cannot add bindings to a locked environment

== 1.5

Check to make sure we are not extracting too many documentation
objects, relative to the stored .result in unit tests.

.names such as .s3method and .overwrite in inner documentation lists
are ignored for find/replacement.

.overwrite signals that all other items of the list should be used
when combining, ignoring previous values stored in those slots.

##sections<< permitted immediately after ###return value

Report which of the users' documentation objects we failed on if the
Parser Function in forall stops with an error.

bugfix: for paren matching heuristic in prefixed.lines Parser
Function, do not consider parens in comments.

extended unit tests: now every testfile is assembled into a package
that will be checked using R CMD check. If there are any warnings,
then we will stop with an error. this poses problems for testfiles
with extensive dependencies and references (such as
scatterplot.R). For these, just set the .dontcheck variable in the
testfile.

== 1.4 unit tests for Parser Functions using inst/testfiles/*.R and
the .result variable in each. we check first if all contents of
.result are correctly extracted, then we check to make sure nothing
more is extracted. This should ensure API stability.

== 1.3 Bug fixes and new code contributed by Ph. Grosjean
  
* examples.after.return is modified to detect where examples start using a
  specific mark (either ##examples<<, or #{{{examples). This way, it does not
  get trap with multiple return() in the function, or the presence of a
  return() in the examples

* examples.in.attr parser is added. It look at an "ex" attribute containing
  code of the examples, either as a character string, or as a function.
  
* Added dependence to 'utils' package (for things like package.skeleton())

* Added the possibility to build a NAMESPACE (argument namespace in
  package.skeleton.dx()). A NAMESPACE is also added to the inlinedocs package.
  
* When DESCRIPTION file is empty, fill required fields with reasonable values,
  e.g., Package: with the same name as dir, Version: 1.0-0, etc.
  
* Automatic detection of S3 methods added, and correct formatting of usage and
  addition of entries in NAMESPACE as required.
  
* Reformatting of special cases usages, like fun<-(x, value) => fun(x) <- value,
  met<-.obj(x, ..., value) => method{met}{obj}(x, ...) <= value, and
  %op%(e1, e2) <- e1 %op% e2
  
* Now the packages in the imports field are considered too, as well for building
  the NAMESPACE file
  
* If there is an 'encoding' field in DESCRIPTION, it is now taken into account
  (according to Writing R Extensions manual, it applies to to R code)!
  
* There is now the possbility to place external examples elsewhere
  that in /tests. Just indicate the subdirectory wher you place them
  in options(inlinedocs.exdir = "exsubsir")

* The names of .Rd files was not computed correctly when functions
  with names like `obj<-`, or `%op%`. Corrected.
  
* The presence of platform specific code in /R/unix or /R/windows is
  not supported by inlinedocs (currnetly). However, this was ignored
  silently. Now, package.skeleton.dx() stops process when it find one
  of these subdirs.
  
* A couple of error messages where badly formatted when they included several
  items like in stop("Need ",names(f)[f]," in ",descfile). Corrected.
  
* Wrong extraction of package names form 'Depends' field in case where packages
  contain one or more dots in their names. Corrected.

* It is necessary to eliminate 'R' from the list of packages to load from
  'Depends' field. Corrected.

* package.skeleton.dx() tried to load packages already laoded, and failed to
  unload packages (including their namespaces, if possible) that where not
  loaded before exiting. Corrected.

== 1.2 new code contributions from thomas, extension mechanism using
Parser Function list

== 1.1 
fixed ... argument bug

== 1.0 

