CHANGES IN knitr VERSION 0.5

NEW FEATURES

• white spaces are allowed before <<>> when using chunk references, and this approach of references is supported in tex and html documents as well
• added a new pattern list named md so that R code blocks can be written more naturally in extended markdown (e.g. GFM and pandoc): use ````{r label, opts}to begin a chunk and ````` (three or more backticks) to begin normal text, and write inline R code in ``r code.here``; it is the default syntax for markdown input, or you can callpat_md()` before `knit()` so knitr can make use of this pattern list to process the input document
• RStudio has added full support to knitr: we can knit HTML and Markdown documents easily now, and markdown output will be automatically converted to an HTML preview page just like TeX to PDF
• if the pattern list is not set in advance, knitr will try to detect the syntax automatically by matching all built-in pattern lists against the input document, e.g. if the input contains <<>>=, the Rnw syntax will be used, and if ````{r}` is detected, the markdown syntax will be used; this feature enables us to use different sets of syntax freely, e.g. we can use Sweave syntax in markdown files and knitr will be able to recognize it (#189)
• new filename extensions with a prefix R or r are introduced: *.Rhtml, *.Rhtm, *.Rtex, *.Rmd and *.Rmarkdown will be recognized as source documents like *.Rnw, and the output filename will be automatically determined by removing the R prefix, e.g. foo.Rmd produces foo.md by default; the old clumsy naming convention foo_knit_.md is still preserved but not recommended any more (#187)
• new function knit2html() to knit an Rmd file (R with markdown) and convert to HTML in one step using the markdown package
• new functions pat_rst() and render_rst() to support reStructuredText; use .. {r label, options} and .. .. to write R code chunks; see https://github.com/yihui/knitr/tree/master/inst/examples/knitr-minimal.Rrst (thanks, Jeffrey Arnold and Ramnath Vaidyanathan)
• new package option self.contained which decides whether to write style definitions (highlighting) in external files or put them in the output document; the highlighting definitions in LaTeX output is often too long, so opts_knit$set(self.contained = FALSE) can help in this case (#176) (thanks, Ramnath Vaidyanathan)
• new package option filter.chunk.end which decides if the chunk.end pattern really means chunk.end; see http://yihui.name/knitr/options (thanks, Joe Cheng)
• syntax highlighting themes are also available to HTML output now; the usage is the same as in LaTeX (#179) (thanks, Ramnath Vaidyanathan)
• the chunk option fig.cap is also used in markdown output now
• the random seed .Random.seed is also cached for the sake of reprodubibility in random simulations
• the function call read_chunk() will be evaluated when tangling R code via purl() (#175) (thanks, Carl Boettiger)
• the default LaTeX output will use the upquote package if it exists so that single quotes are straight in the output (thanks, Mathematical Coffee http://bit.ly/IKjluw)
• the chunk option engine is back but it is used to run code from other languages instead of just ignoring the code in chunks which have engine != 'R'; currently this option is still in rough edges and only supports python and awk; other languages may be added in future versions, but users can also do it by themselves by knit_engines$set(language = function(options) {...}); see an example at system.file('examples', 'knitr-lang.Rmd') (#201)
• new function write_bib() to write Bibtex citations for R packages; see the main manual for a sample usage (#13)
• hook_pdfcrop() also supports cropping other image formats like PNG or JPEG through ImageMagick (convert -trim) (#203) (thanks, @r2d3)

MAJOR CHANGES

• knitr will completely stop when duplicated labels are found and the corresponding chunks are non-empty; in previous version, only a warning is given and R code in later chunks will override previous chunks (#185) (thanks, Randall Pruim)
• the default graphical device for HTML and markdown output is png() now to avoid the possible unexpected consequence that PDF plots are included in an HTML page which will not work normally in browsers
• markdown output will use the extended markdown hooks by default now: render_markdown(strict = FALSE); in the past the default was render_jekyll(); the major difference is that the code blocks are put in r and ; if you want the strict markdown syntax, you can all render_markdown(strict = TRUE) which will indent code blocks by 4 spaces
• render_gfm() has been removed because the name can be misleading; the main purpose of this function was to put code blocks in `````, and now we can replace it by render_markdown(FALSE); other markdown flavors also support such fenced code blocks (e.g. pandoc) -- it is not limited to Github only
• the default value for the fig.path option is 'figure/' now so that plots will be put under this directory by default; the default was './' in the past which makes the current directory messy when there are many plots
• knitr will fully stop when an error is encountered in knit(); in the past only a message was issued in this case in an interactive R session
• the package option all.patterns has been dropped; please use the objects all_patterns or knit_patterns directly if you want to tweak the syntax

BUG FIXES

• the compilation of tikz graphics can hang up when there are TeX errors in the tikz file; now we use \nonstopmode to avoid hiccup (#188)
• multiple devices per chunk was buggy (#181)
• S4 objects will be printed by show() instead of print(); this is a bug of the evaluate package, which has been fixed (please update it to be > 0.4.2)

MISC

• it is recommended to use opts_chunk$set() to set global chunk options now instead of \SweaveOpts{}; all documentation has been updated (#216)
• number of downloads (https://github.com/yihui/knitr/downloads) of knitr documentation before I removed and updated them on GitHub: c(main = ?, graphics = 275+)