NEWS

---

Statamarkdown 1.1.0

• purl_stata() now writes each chunk's header in the *%% label, options ---- form which knitr uses when it tangles a script in another language, instead of its own * ---- label ---- form, so a do-file written by purl_stata() and one tangled by knitr::purl() look the same. The engine is dropped from the header, since the do-file implies it.
• purl_stata() now uses knitr::partition_chunk() to split a chunk's option comments from its code, so they are recognised and parsed exactly as knitr does when the document is knitted: *| comments in a stata chunk and #| comments in a chunk using the older engine='stata' form. Values such as purl: no are therefore understood, and a //| line, which knitr does not treat as an option comment, is now kept as (Stata comment) code instead of being dropped.
• spinstata() no longer needs engine='stata' in the chunk headers of a "do" file: code in a chunk is taken to be Stata code unless it is marked as R code with the /*R ... R*/ markup, or the chunk header sets an engine option itself. With knitr >= 1.53, Stata is also used for code which is not preceded by a chunk header at all, in a document which has no chunks of R code.
• spinstata() now calls knitr::spin() directly, instead of the modified copy of it which the package had carried since knitr 1.22. The copy was needed because spin() could not handle a script which is not valid R code, which was fixed in knitr 1.46 (yihui/knitr#1773). As a result spinstata() gains the features spin() has since acquired, including the format = "qmd" output format and the # %% and #| chunk delimiters, and the package now requires knitr >= 1.46.
• find_stata(message = FALSE) is now silent when no Stata executable is found; previously the "No Stata executable found." message was printed whatever the value of message. The help file examples use this, so the reference pages of the pkgdown site, which is built without Stata, no longer end with that message.

Statamarkdown 1.0.0 (2026-08-21)

• Rendering several documents in one R session now works: knitr restores the chunk options and knit hooks after each document, which left the second document without the Stata executable path (failing with "error in running command") and without the collectcode hook. The engine now falls back to the executable located when the package was attached, and re-registers the hook, so it is no longer necessary to detach("package:Statamarkdown") between documents.
• New chunk option stata.fig=TRUE: exports the graph drawn by a Stata chunk and includes it in the output document. The figure goes through knitr's standard plot machinery, so fig.cap, fig.alt (alternative text, for accessibility), out.width, out.height, fig.align, fig.link and fig.path all work as for R chunks. The export format is set with stata.fig.format (default "svg", which Stata batch mode supports on all platforms). In Rmd and Quarto documents the hyphenated spellings stata-fig and stata-fig-format are also accepted when using YAML style chunk options. (thanks @kylebutts in #28)
• New function purl_stata(), the Stata analogue of knitr::purl(): extracts the code from the Stata chunks of an R Markdown or Quarto document into a do-file, recording each chunk's header (label and options) as a Stata comment, and following knitr::purl()'s documentation levels (2 also carries the document text over as Stata comments). Chunks with purl=FALSE or eval=FALSE are skipped, including via option comments in any of the forms knitr accepts in Stata chunks (#|, *|, //|). (thanks @davidaarmstrong in #39)
• Fixed the collectcode chunk option when knitr evaluates chunks in a different directory to the document, for example for a document in a subfolder of an RStudio project with the "Evaluate chunks in directory: Project" option set, or when knitting with rmarkdown::render(knit_root_dir=). The temporary profile.do is now created, and any pre-existing profile.do snapshot and restored, in the directory where the chunks (and therefore Stata) run. (thanks @DaniMori in #17)
• Switch from including the html docs files to including them within the package as precomputed vignettes using Quarto.
• Tweak helpfile examples such that they run within RStudio.
• Add markdown to Suggests dependencies because it is needed for a code path within spinstata (in fact the one in that helpfile).
• Guard two of the helpfile examples for the presence of the rmarkdown package.
• Only activate nocommands, nooutput and quietly when set to TRUE
• Add some tests
• Convert documentation to use roxygen2

Statamarkdown 0.9.7 (2026-07-19)

• Fixed loss of all chunk output when the Stata log does not contain an "end of do-file" line.
• spinstata() no longer overwrites its input file; a statafile without a .do extension is now an error, and the extension check is case-insensitive.
• Fixed a crash in the collectcode chunk option when engine.path is given as a plain string rather than a list.
• The cleanup of the temporary profile.do created by collectcode now locates the knitr::knit() call on the call stack instead of assuming a hardcoded stack depth, making it robust to changes in knitr's internals, child documents, and different ways of invoking knitr.
• cleanlog=TRUE now removes all lines of multi-line loop commands and wrapped (continued) command lines from the output.
• cleanlog=TRUE no longer deletes genuine output lines that begin with a decimal point.
• Fixed spinstata() block tracking when a marked-up block starts and ends on the same line.
• Fixed an error in spinstata() on R >= 4.3 when producing Rnw or Rtex output.
• find_stata() on Windows now stops searching as soon as an executable is found, and the non-existent "StataNow19" directory stub has been removed from the search.
• Removed a duplicated "No Stata executable found" startup message when attaching the package without Stata installed.
• Minor performance improvements: the Stata executable is located once at package load, and per-chunk processing avoids repeated system calls and redundant vector allocations.
• Change of maintainer to Tom Palmer. Many thanks to Doug Hemken for creating and maintaining this amazing package.
• Add the precomputed vignettes to the distributed CRAN package

Statamarkdown 0.9.6 (2025-10-07)

• find_stata() now searches /Applications/StataNow on macOS, the StataNow directories on Windows, and two additional installation directories on Linux.
• Fixed bugs in the Windows directory search.
• Fixed help file cross-reference links for the new CRAN standard.

Statamarkdown 0.9.4 (2025-08-23)

• Updated find_stata() for Stata 19 and the SSCC installation path.

Statamarkdown 0.9.3

• GitHub-only release, not published on CRAN.
• Updated find_stata() for Stata 19.

Statamarkdown 0.9.2 (2023-12-04)

• Cleaner Stata log output, and the log cleanup now accounts for unset chunk options.
• Fixed the collectcode chunk option for knitr >= 1.45.
• Package setup switched from .onAttach() to .onLoad().
• Cleaned up package dependencies and other CRAN submission issues.

Statamarkdown 0.8.0 (2023-06-01)

• Thoroughly reworked engine output processing and package loading.
• Edited the vignettes.

Statamarkdown 0.7.4 (2023-05-25)

• Fixed handling of Stata executable paths containing spaces on macOS and Unix.
• Updated the required knitr version.

Statamarkdown 0.7.3

• GitHub-only release, not published on CRAN.
• Updated find_stata() for Stata 18.
• Improved the help file examples, in particular for spinstata().
• Changed the vignette rendering, and pointed the README to CRAN.

Statamarkdown 0.7.2 (2023-02-15)

• Better removal of the "Running .../profile.do" line from the log for long installation paths.
• Added further Unix installation paths to find_stata(), suggested by Tom Palmer.

Statamarkdown 0.7.1

• GitHub-only release, not published on CRAN.
• Improved the find_stata() directory search on Linux, using Sys.which().

Statamarkdown 0.7.0

• GitHub-only release, not published on CRAN.
• Updated find_stata() for Stata 17, with macOS fixes.
• Fixed a problem with factor variable names.

Statamarkdown 0.6.0

• GitHub-only release, not published on CRAN.
• Cleaned up messages in "notebook" mode.
• Fixed a macOS path problem, and added StataIC to the macOS search.
• Fixed log cleanup of the "Running ..." line for Stata 16, and the eval chunk option.
• Prebuilt the vignettes for the binary version of the package.
• Added Philipp Lepert as a contributor.

Statamarkdown 0.4.4

• GitHub-only release, not published on CRAN.
• First tagged release: a knitr language engine for Stata with cleaned log output, the collectcode chunk option for linking code blocks, spinstata() for spinning specially marked-up Stata do-files into reports, and find_stata() to locate the Stata executable on Windows, macOS, and Linux.