Compare Strings with or without Collation
These functions may be used to determine if two strings are equal, canonically equivalent (this is performed in a much more clever fashion than when testing for equality), or to check whether they are in a specific lexicographic order.
stri_compare(e1, e2, ..., opts_collator = NULL) stri_cmp(e1, e2, ..., opts_collator = NULL) stri_cmp_eq(e1, e2) stri_cmp_neq(e1, e2) stri_cmp_equiv(e1, e2, ..., opts_collator = NULL) stri_cmp_nequiv(e1, e2, ..., opts_collator = NULL) stri_cmp_lt(e1, e2, ..., opts_collator = NULL) stri_cmp_gt(e1, e2, ..., opts_collator = NULL) stri_cmp_le(e1, e2, ..., opts_collator = NULL) stri_cmp_ge(e1, e2, ..., opts_collator = NULL)
e1, e2 |
character vectors or objects coercible to character vectors |
... |
additional settings for |
opts_collator |
a named list with ICU Collator's options,
see |
All the functions listed here are vectorized over e1
and e2
.
stri_cmp_eq
tests whether two corresponding strings
consist of exactly the same code points, while stri_cmp_neq
allows
to check whether there is any difference between them. These are
locale-independent operations: for natural language processing,
where the notion of canonical equivalence is more valid, this might
not be exactly what you are looking for, see Examples.
Please note that stringi always silently removes UTF-8
BOMs from input strings, therefore, e.g., stri_cmp_eq
does not take
BOMs into account while comparing strings.
stri_cmp_equiv
tests for canonical equivalence of two strings
and is locale-dependent. Additionally, the ICU's Collator may be
tuned up so that, e.g., the comparison is case-insensitive.
To test whether two strings are not canonically equivalent,
call stri_cmp_nequiv
.
stri_cmp_le
tests whether
the elements in the first vector are less than or equal to
the corresponding elements in the second vector,
stri_cmp_ge
tests whether they are greater or equal,
stri_cmp_lt
if less, and stri_cmp_gt
if greater,
see also, e.g., %s<%
.
stri_compare
is an alias to stri_cmp
. They both
perform exactly the same locale-dependent operation.
Both functions provide a C library's strcmp()
look-and-feel,
see Value for details.
For more information on ICU's Collator and how to tune its settings
refer to stri_opts_collator
.
Note that different locale settings may lead to different results
(see the examples below).
The stri_cmp
and stri_compare
functions
return an integer vector representing the comparison results:
-1
if e1[...] < e2[...]
,
0
if they are canonically equivalent, and 1
if greater.
All the other functions return a logical vector that indicates
whether a given relation holds between two corresponding elements
in e1
and e2
.
Collation - ICU User Guide, http://userguide.icu-project.org/collation
Other locale_sensitive:
%s<%()
,
about_locale
,
about_search_boundaries
,
about_search_coll
,
stri_count_boundaries()
,
stri_duplicated()
,
stri_enc_detect2()
,
stri_extract_all_boundaries()
,
stri_locate_all_boundaries()
,
stri_opts_collator()
,
stri_order()
,
stri_rank()
,
stri_sort_key()
,
stri_sort()
,
stri_split_boundaries()
,
stri_trans_tolower()
,
stri_unique()
,
stri_wrap()
# in Polish, ch < h: stri_cmp_lt('hladny', 'chladny', locale='pl_PL') # in Slovak, ch > h: stri_cmp_lt('hladny', 'chladny', locale='sk_SK') # < or > (depends on locale): stri_cmp('hladny', 'chladny') # ignore case differences: stri_cmp_equiv('hladny', 'HLADNY', strength=2) # also ignore diacritical differences: stri_cmp_equiv('hladn\u00FD', 'hladny', strength=1, locale='sk_SK') marios <- c('Mario', 'mario', 'M\\u00e1rio', 'm\\u00e1rio') stri_cmp_equiv(marios, 'mario', case_level=TRUE, strength=2L) stri_cmp_equiv(marios, 'mario', case_level=TRUE, strength=1L) stri_cmp_equiv(marios, 'mario', strength=1L) stri_cmp_equiv(marios, 'mario', strength=2L) # non-Unicode-normalized vs normalized string: stri_cmp_equiv(stri_trans_nfkd('\u0105'), '\u105') # note the difference: stri_cmp_eq(stri_trans_nfkd('\u0105'), '\u105') # ligatures: stri_cmp_equiv('\ufb00', 'ff', strength=2) # phonebook collation stri_cmp_equiv('G\u00e4rtner', 'Gaertner', locale='de_DE@collation=phonebook', strength=1L) stri_cmp_equiv('G\u00e4rtner', 'Gaertner', locale='de_DE', strength=1L)
Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.