NEWS.md
profmem() gained argument on_error to control what should happen when there is a run-time error during profiling. Known actions are "error", "warning", and "ignore" (default).
print() for Rprofmem objects now report on any run-time errors that occurred while profiling the R expression.
print() for Rprofmem objects gained logical argument `calls’ to control whether calls (“traces”) are outputted.
expr and newpage of print() for Rprofmem can now be set via options profmem.print.expr and profmem.print.newpage.Added documentation on what an ‘Rprofmem’ data.frame contains.
Added example for readRprofmem().
The bytes column returned by readRprofmem() would be of type integer when zero allocation events where read, whereas otherwise it would be of type numeric.
print(x) for Rprofmem now uses an explicit stringsAsFactors = FALSE when internally coercing x to a data frame.
Added profmem_begin() and profmem_end() as an alternative to profmem().
Added profmem_suspend() and profmem_resume() for suspending and resuming an active profmem session.
Added profmem_status() for checking whether profmem is inactive, active, or suspended.
Added profmem_depth() to get the current depth of nested profmem sessions.
print() for Rprofmem now reports on the threshold used, if known.
print() for Rprofmem gained arguments expr = TRUE and newpage = FALSE, controlling whether the profiled R expression and ‘new page’ entries should be displayed or not.
Added option profmem.threshold for controlling the default threshold level for profmem() and profmem_begin().
Vignette: Updated the example in Section ‘An example where memory profiling can make a difference’ due to improvements in R (>= 3.4.0), which made the previous example no longer valid.
help("profmem") how provide much more details previously only available in the vignette.
readRprofmem() failed to “workaround” another buggy Rprofmem() output.readRprofmem().Added profmem().
Package created.