Parsing Expression Grammars

Parsing Expression Grammars, or PEGs, describe a formal language about how to recognize strings in a set of text. For example, capture any letter:

plus< alpha >

Combine to form a grammar, e.g., must match name, one comma, a space, then match numbers.

must< name, one< ',' >, space, numbers, eof >

pegax package

Get it on GitHub: https://github.com/ropenscilabs/pegax

PEG work done in C++ using PEGTL (https://github.com/taocpp/PEGTL) via the piton (https://github.com/Ironholds/piton) R package.

pegax package

Installation

remotes::install_github("ropenscilabs/pegax")
library(pegax)

Examples

# Author name
pgx_authority_names("Linnaeus, 1758")
#> [1] "Linnaeus"

# Author year
pgx_authority_years("Linnaeus, 1758")
#> [1] "1758"

# Parse ranks
pgx_ranks("Fagus sylvatica subsp. orientalis")
#> [1] "subsp."

# Many ranks at once
pgx_ranks(c("Helianthus annuus var. annuus",
   "Helianthus annuus ssp. annuus", 
   "Caulerpa cupressoides forma nuda"))
#> [1] "var."   "ssp."   "forma."

# Scientific names w/o authorities
nms <- c("Fagus sylvatica subsp. orientalis",
  "Potamogeton iilinoensis var. ventanicola",
  "Callideriphus flavicollis morph. reductus",
  "Chlorocyperus glaber form. fascicula",
  "Sphaerotheca fuliginea f. dahliae")
dplyr::bind_rows(lapply(nms, pgx_sciname))
#>          genus     epithet  rank   infraspecific
#>          Fagus   sylvatica  subsp.    orientalis
#>    Potamogeton iilinoensis  var.     ventanicola
#>  Callideriphus flavicollis  morph.      reductus
#>  Chlorocyperus      glaber  form.      fascicula
#>   Sphaerotheca   fuliginea  f.           dahliae

benchmark: it’s pretty fast

library(microbenchmark)
library(charlatan)
per <- charlatan::PersonProvider$new()
date <- charlatan::DateTimeProvider$new()
# 10,000 name/year strings, e.g., Lueilwitz, 1945
x <- replicate(10^4, paste( 
    sample(per$person$last_names, 1),
    date$year(),
    sep = ", "))
# years
system.time(pgx_authority_years(x)) # 0.025 sec
# names
system.time(pgx_authority_names(x)) # 0.023 sec

future work

• Unicode support
• Full name parsing (names, ranks, annotations, authorities)
• R DSL to make custom PEGS for taxonomy

pegax(plus(alpha), one(,), numbers(,))

Use cases

• Researchers/etc. use package directly to parse their own set of names
• R developers import to leverage name parsing for domain specific use case
• If/when DSL for custom name parsing: cover all corner cases

contact: scott@ropensci.org

icons from FontAwesome v5.0.13 fontawesome.com