Skip to contents

Record provenance from code that does not use apply_fit()

Source: R/fit_provenance.R
record_provenance.Rd

apply_fit and apply_fits record their recipes themselves. Code that writes results another way uses these functions to add to the same metadata(se)$provenance entry.

Usage

record_fit_provenance(
  se,
  assay,
  fit_source,
  fn,
  fit_args = list(),
  input_assay = NULL,
  replace = FALSE
)

record_processing_step(
  se,
  step,
  settings = list(),
  fns = list(),
  max_bytes = .provenance_max_value_bytes
)

Arguments

se

SummarizedExperiment

assay

string; assay the function wrote.

fit_source

string; fit_source of the written rows. Assays without a fit_source column use it only as a key.

fn

function of the whole SE that wrote assay, called as fn(se, ...).

fit_args

named list of the arguments fn was called with, other than se.

input_assay

string or NULL; assay the results were computed from, hashed to tell later whether a re-run used the same data.

replace

logical; TRUE when fn overwrote the whole assay, so the recipes of other fit sources in it no longer apply.

step

string; name of the processing step.

settings

named list of plain-data settings of the step.

fns

named list of the functions the step used, e.g. list(rate_fn = rate_fn).

max_bytes

numeric; largest size in bytes of one setting that is stored. Larger settings are listed in settings_not_stored. Raise it for settings that are large by nature, such as the list of input files.

Value

se with the provenance entry updated.

Details

record_fit_provenance() is for a function of the whole SE that writes an assay, e.g. a combination scoring step. The recipe keeps the function and its arguments, so refit_from_provenance can call it again as fn(se, ...). Functions passed among the arguments (e.g. a score function) are stored as function records.

record_processing_step() is for a step whose result cannot be recomputed from the object it returns, e.g. growth rates computed from an assay that the output no longer carries. It stores the functions and settings of the step under provenance$steps[[step]].

Examples

mae <- get_synthetic_data("finalMAE_small.qs2")
se <- mae[["single-agent"]]
se <- record_processing_step(
  se, "growth_rates",
  settings = list(periods = list(early = c(0, 48))),
  fns = list(rate_fn = function(dt) 0)
)
names(get_SE_provenance(se)$steps)
#> [1] "growth_rates"