Build a file-to-target-path plan for a Psych-DS-style layout

Maps every file data_check found to the location it would occupy in a Psych-DS-shaped layout: data files under data/, code under analysis/, materials under materials/, documentation under documentation/ (or documentation/codebooks/ for codebooks), the root readme/licence at the collection root, and anything unclassified under unknown/. This is pure file/folder organisation – it does not check Psych-DS compliance (that is psychds_check()) and it does not write anything to disk.

Usage

psychds_file_plan(structure_df, group_no_evidence = FALSE)

Arguments

structure_df: the structure table from data_check (one row per repository file, with at least file_name/file_path/data_type; the group/doc_role/referenced_by columns are used when present).
group_no_evidence: TRUE when no file path, repository split, or LLM answer named a study anywhere in this repository, so every file fell back to a single default study (data_check's own group_no_evidence output). Carried through to the return value only; does not affect the plan.

Value

a list with table (one row per file: file_name, data_type, group, current_path, target_path, status – "present"/"move"/ "excluded" –, convert, original_target, referenced_by) and group_no_evidence (passed through unchanged). table is an empty data.frame when structure_df is NULL or has no rows.

Details

data_check assigns every file except the collection-level root README/ ro-crate-metadata.json to exactly one study group – deterministically where path/repository/code-reference evidence allows, and via an LLM only for the residual cases (see data_group_llm()). A multi-study repository is modelled with a study-<group>/ directory per study (each a complete Psych-DS dataset); only the root readme and licence sit at the collection root beside them.

A tabular data file whose source is not already .csv (.xlsx/.sav/ .dta/...) is planned to be converted to .csv at its target path, with the untouched original kept alongside it (convert/original_target columns) – no conversion is actually performed here, only planned. A raw (non-tabular) data file (.npy/.h5/...) keeps its own real name instead of a _data.csv target, since it cannot be read as a table.

This is a pure function of structure_df: it takes no paper and makes no data_check/module_run() call of its own, so callers resolve their own structure_df (chained output, a saved build, or a fresh data_check run) and pass it in directly – the same frame-dependent get_prev_outputs() chain lookup every module relies on only works correctly one call deep from module_run()'s own eval(), not from a second, ordinary function call nested inside it.

Examples

paths <- list.files(demofile("repo"), recursive = TRUE)
structure_df <- data.frame(file_name = basename(paths), file_path = paths,
                           data_type = data_classify_files(basename(paths)))
psychds_file_plan(structure_df)$table[, c("file_name", "status", "target_path")]
#>                   file_name  status                            target_path
#> 1                   LICENSE    move                  documentation/LICENSE
#> 2                 README.md    move                documentation/README.md
#> 3              analysis.qmd    move                  analysis/analysis.qmd
#> 4              clean_data.R    move                  analysis/clean_data.R
#> 5            references.bib    move           documentation/references.bib
#> 6            stroop_raw.csv    move      data/raw/study-stroopraw_data.csv
#> 7           stroop_data.csv    move         data/study-stroopdata_data.csv
#> 8  dataset_description.json    move          data/dataset_description.json
#> 9  stroop_data_codebook.csv present documentation/stroop_data_codebook.csv
#> 10                renv.lock    move                     analysis/renv.lock