Create a hot-reloading data store

Registers named datasets backed by files on disk and hands them to route handlers without globals or <<-. Each aurora_data_get() checks the file's modification time and transparently re-reads it if an external process (e.g. an ETL job) has rewritten it – the stateless equivalent of the reference app's carregar_bases() mtime trick.

Usage

aurora_data_store(..., dir = ".", readers = list())

Arguments

...

Named file paths to register (e.g. sales = "data/sales.rds"). The reader is inferred from the file extension.

dir

Base directory that relative dataset paths are resolved against. Resolved to an absolute path once, when the store is created (not at read time), so later changes to the working directory cannot break reads. Defaults to "."; since the store is normally created while a helpers/*.R file is sourced (cwd = app root), relative paths like "data/x.rds" anchor to the app root. Absolute dataset paths are used as-is regardless of dir.

readers

Named list of reader functions keyed by lowercase file extension, merged over (and overriding) the built-ins (rds, csv, parquet).

Value

An object of class aurora_data_store.

Details

Define the store once in a helpers/*.R file (sourced before routers are parsed) and read from it in handlers:

# helpers/data.R
store <- aurora_data_store(sales = "data/sales.rds", dir = ".")

# routers/sales.R
#* @get /api/sales
#* @serializer json
function() aurora_data_get(store, "sales")

Examples

f <- tempfile(fileext = ".rds")
saveRDS(data.frame(x = 1:3), f)
store <- aurora_data_store(demo = f)
aurora_data_get(store, "demo")
#>   x
#> 1 1
#> 2 2
#> 3 3