[foersterst]

tp_run() used to return a list built directly from the output JSON via jsonlite. That list was long and hard to read, and it always needed a tp_parse…() call afterward to become usable. This PR changes that. Now, tp_run() returns a tidy data frame produced by tp_parse_smc() or tp_parse_mcmc(), called internally for any model those two functions support (models that return trees still need dedicated parsers, which we haven't built yet).

tp_run() picks the right parser automatically, based on the inference method and model passed to tp_compile(). It reads that information from the sampler and decides which parser to call. For models without a parser yet, it returns the path to the output JSON along with a message instead.

The PR also updates the documentation for all functions involved.

---

TreePPL Release Notes

Added

• At the end of a run, tp_run() now prints a summary with runtime information and output directories/files.
• New argument wide in tp_parse_smc() and tp_parse_mcmc() to choose between wide (standard in other software) and long (good for plotting) format. Default = TRUE.

Changed

• tp_run() now returns a tidy data frame (in wide format by default) instead of a list. It automatically calls tp_parse_smc() or tp_parse_mcmc(), depending on the inference method and model passed to tp_compile(). For models without a parser yet (e.g., models that return trees), it returns the path to the output JSON file along with a message.
• tp_parse_smc() and tp_parse_mcmc() now take a character vector with the path(s) to the JSON output file(s) instead of the list returned by tp_run().
• tp_mcmc_convergence() and tp_smc_convergence() now accept data frames in either long or wide format.

Deprecated

Removed

Fixed

Security