TODOs

Examples: RcppExports fun/arg parsing.

capture_melt_multiple(fill=TRUE) for adding NA entries for missing
columns? are there any compelling use cases?

Should we try to eliminate non-capturing groups when they are not necessary?
e.g. below the only thing in the inside non-capturing group is a capturing group,
so the inside non-capturing group could be removed.
> nc::var_args_list(nc::quantifier(kit=nc::alternatives("PP16", "IP", "P"), "?"))$pattern
[1] "(?:(?:(PP16|IP|P))?)"

2020.5.13

tidyr::pivot_longer examples use names_transform (not present in tidyr
1.0.3) instead of names_ptypes which gives an error now for some
reason. https://github.com/tidyverse/tidyr/pull/942/files

2020.3.25

fix typo in capture_melt_single example.

stop_for_engine tells user to get re2r from github.

2020.3.23

reduce repetition in tests, test_engines defined in a separate
inst/test_engines.R file.

2020.2.27

PROVEDit file names example in capture first vignette.

alternatives(pattern1, pattern2, ...) generates pattern1|pattern2,
with support for named arguments.

2020.2.24

capture_melt_multiple bugfix when "count" is one of the columns to
output.

2020.1.16

test for pattern with literal groups.

2019.12.12

time command line example for capture_all_str.

more informative error message for capture_first_df.

2019.11.22

quantifier.

Do not run emoji examples on Solaris.

2019.10.19

dot prefix now ok for capture_melt_multiple.

informative error messages to enforce unique column names in input and
output of capture_melt_*.

2019.10.14

capture_first_melt -> capture_melt_single.

capture_first_melt_multiple -> capture_melt_multiple.

2019.10.11

new vignettes 3=melt 4=comparison.

move unicode data from example to extdata in order to avoid the
following on CRAN solaris:

    Warning in deparse(e[[2L]]) :
     it is not known that wchar_t is Unicode on this platform

2019.9.28

capture_first_melt_multiple inspired by data.table::patterns.

capture_first_melt inspired by tidyr::pivot_longer.

Suggest tidyr for who data example.

2019.9.18

additional citation/medline web page data/examples.

2019.9.16

do not do engine tests (with rename library dirs) on CRAN.

remove print/str from error messages for CRAN.

group, field functions.

2019.9.12

always output data.table.

2019.9.10

test 0 column result via NULL names and rownames, not
identical/expected data.frame.

2019.9.4

engine error message includes install.packages.

2019.9.3

dput used for did you mean error message.

vignettes separated, 1=first 2=all.

2019.8.22

DESCRIPTION updated.

Informative error for named character string patterns.

2019.8.21

main functions renamed to capture_* for auto-complete convenience.

on https://github.com/gagolews/stringi/issues/279 it says ICU >= 59 is
required for emoji support. but travis uses Ubuntu xenial system ICU
(55.1) by default. Ubuntu bionic has ICU 60.2
https://launchpad.net/ubuntu/bionic/+source/icu so travis build will
most likely be fixed by telling travis to use bionic. Actually dist:
bionic builds use xenial still for language: r,
https://travis-ci.community/t/for-dist-bionic-xenial-vm-is-used-instead-with-language-rust/4487/2
So instead we fix the build by making if interactive() for this
example, and adding another engine="ICU" example based on
http://userguide.icu-project.org/strings/regexp for str_capture_all.

docs.

vignette.

FIXED: needed to escape backslashes in examples.
Executing example(vec_capture_first) gives
> vec_capture_first(
+ "a\U0001F60E#",
+ emoji="\p{EMOJI_Presentation}",
Erreur : '\p' est un code escape non reconnu dans une chaîne de caractères débutant ""\p"
(but executing the example from C-c C-c in R/vec_capture_first.R works fine)

2019.8.14

first version forked from namedCapture.

vec_capture_first, str_capture_all, df_capture_first pass tests for
three engines: ICU, PCRE, RE2.