NEWS

---

rfsrs 0.3.2

Breaking changes

• Direct Rust bindings are now internal. Previously-exported fsrs_default_parameters(), fsrs_initial_state(), fsrs_next_state(), fsrs_next_interval(), fsrs_retrievability() are renamed with a _raw suffix and no longer in the NAMESPACE. Use the validated wrappers instead: fsrs_parameters(), fsrs_new_card_state(), fsrs_next_memory_state(), fsrs_interval(), fsrs_recall_probability(). Direct callers can still reach the bindings via rfsrs:::fsrs_*_raw, but these are not a stable API.

Documentation

• README Low-Level table reduced to the validated wrappers only.

rfsrs 0.3.1

Bug Fixes

• fsrs_next_memory_state() / fsrs_repeat(): fractional elapsed_days now rounds at the Rust FFI boundary instead of truncating. elapsed_days = 0.7 used to collapse to 0, giving R = 1.
• fsrs_anki_to_reviews(): accepts rid and review_type column names used by ankiR::anki_revlog().
• fsrs_version(): returns "FSRS-6". Default parameters have always used FSRS6_DEFAULT_DECAY = 0.1542 (fsrs-rs 5.2.0 DEFAULT_PARAMETERS[20]).
• fsrs_evaluate() / fsrs_optimize(): skip items whose current review has delta_t == 0. R = 1 exactly makes log(1 - R) diverge, which returned log_loss = NaN.

Documentation

• README: Anki import example now uses anki_revlog() instead of the non-existent read_anki().
• README: forgetting-curve formula matches fsrs-rs: R(t) = (1 + factor · t/S)^(-decay), factor = 0.9^(1/-decay) - 1.
• README: Low-Level Functions table split into validated R wrappers and direct Rust bindings.

rfsrs 0.1.0

• Initial release
• R bindings for fsrs-rs Rust library via rextendr
• Core functions:
  • fsrs_default_parameters(): Get FSRS-6 default parameters (21 values)
  • fsrs_initial_state(): Create initial memory state for new cards
  • fsrs_next_state(): Calculate memory state after a review
  • fsrs_next_interval(): Get optimal interval for next review
  • fsrs_retrievability(): Calculate probability of recall
• Supports all four FSRS ratings (Again, Hard, Good, Easy)
• Available on r-universe: https://chrislongros.r-universe.dev
• Builds successfully on Linux and macOS (Windows builds pending)