Input and Output#

This tutorial is based on the paper on stringi that was published in the Journal of Statistical Software; see [2]. To learn more about R, check out Marek’s open-access (free!) textbook Deep R Programming [3].

This section deals with some more advanced topics related to the operability of text processing applications between different platforms. In particular, we discuss how to ensure that data read from various input connections are interpreted correctly.

Dealing with Unicode Code Points#

The Unicode Standard (as well as the Universal Coded Character Set, i.e., ISO/IEC 10646) currently defines over 140,000 abstract characters together with their corresponding code points – integers between 0 and 1,114,111 (or 0000 and 10FFFF in hexadecimal notation, see In particular, here is the number of the code points in some popular categories (compare Matching Individual Characters), such as letters, numbers, and the like.

z <- c("\\p{L}", "\\p{Ll}", "\\p{Lu}", "\\p{N}", "\\p{P}", "\\p{S}",
  "\\w", "\\d", "\\s")
  setdiff(1:0x10ffff, c(0xd800:0xf8ff))), z), names=z)
##  \\p{L} \\p{Ll} \\p{Lu}  \\p{N}  \\p{P}  \\p{S}     \\w     \\d     \\s 
##  131241    2155    1791    1781     798    7564  134564     650      25

Yet, most of the code points are still unallocated. The Unicode standard is occasionally updated, e.g., the most recent versions were supplemented with over 1,000 emojis.

The first 255 code points are identical to the ones defined by ISO/IEC 8859-1 (ISO Latin-1; “Western European”), which itself extends US-ASCII (codes ≤ U+007f). For instance, the code point that we are used to denoting as U+007A (the “U+” prefix is followed by a sequence of hexadecimal digits; U+007A corresponds to decimal 122) encodes the lower case letter “z”. To input such a code point in R, we write:

"\u007A"  # or "\U0000007A"
## [1] "z"

For communicating with ICU and other libraries, we may need to escape a given string, for example, as follows (recall that to input a backslash in R, we must escape it with another backslash).

x <- "zß你好"
## [1] "z\\u00df\\u4f60\\u597d"

Even though some output devices might be unable to display certain code points correctly (due to, e.g., missing fonts), the correctness of their processing with stringi is still guaranteed by ICU.

Character Encodings#

When storing strings in RAM or on the disk, we need to decide upon the actual way of representing the code points as sequences of bytes. The two most popular encodings in the Unicode family are UTF-8 and UTF-16:

x <- "abz0ąß你好!"
stri_encode(x, to="UTF-8", to_raw=TRUE)[[1]]
##  [1] 61 62 7a 30 c4 85 c3 9f e4 bd a0 e5 a5 bd 21
stri_encode(x, to="UTF-16LE", to_raw=TRUE)[[1]]  # little-endian
##  [1] 61 00 62 00 7a 00 30 00 05 01 df 00 60 4f 7d 59 21 00

R’s current platform-default encoding, which we’ll refer to as the native encoding, is defined via the LC_CTYPE locale category in Sys.getlocale(). This is the representation assumed, e.g., when reading data from the standard input or from files (e.g., when scan() is called). For instance, Central European versions of Windows will assume the “windows-1250” code page. On the other hand, MacOS as well as most Linux boxes work with UTF-8 by default[1].

All strings in R have an associated encoding mark which can be read by calling Encoding() or, more conveniently, stri_enc_mark(). Most importantly, strings in ASCII, ISO-8859-1 (”latin1”), UTF-8, and the native encoding can coexist. Whenever a non-Unicode string is passed to a function in stringi, it is silently converted to UTF-8 or UTF-16, depending on the requested operation (some ICU services are only available for UTF-16 data). Over the years, this has proven a robust, efficient, and maximally portable design choice – Unicode can be thought of as a superset of every other encoding. Moreover, in order to guarantee the correctness and high performance of the string processing pipelines, stringi always[2] outputs UTF-8 data.

Reading and Writing Text Files#

According to a report by W3Techs, as of 2021-09-28, 97.3% of websites use UTF-8. Nevertheless, other encodings can still be encountered.

If we know the encoding of a text file in advance, stri_read_lines() can be used to read the data in a manner similar to the built-in readLines() function (but with a much easier access to encoding conversion):

For instance, below we read a text file (see encoded in ISO-8859-1:

x <- stri_read_lines("ES_latin1.txt", encoding="ISO-8859-1")
head(x, 4)  # x is in UTF-8 now
## [1] "CANTO DE CALÍOPE - Miguel de Cervantes"
## [2] ""                                      
## [3] "Al dulce son de mi templada lira,"     
## [4] "prestad, pastores, el oído atento:"

We can call stri_write_lines() to write the contents of a character vector to a file (each string will constitute a separate text line), with any output encoding.

Detecting Encodings#

If a file’s encoding is not known in advance, there are a certain functions that can aid in encoding detection. First, we can read the resource in the form of a raw-type vector:

x <- stri_read_raw("ES_latin1.txt")
head(x, 24)  # vector of type raw
##  [1] 43 41 4e 54 4f 20 44 45 20 43 41 4c cd 4f 50 45 20 2d 20 4d 69 67 75 65

Then, to guess the encoding, we can call, e.g.:

## [1] FALSE
stri_enc_isutf8(x)   # false positives are possible
## [1] FALSE

Alternatively, we can use:

stri_enc_detect(x)  # based on heuristics
## [[1]]
##     Encoding Language Confidence
## 1 ISO-8859-1       es       0.81
## 2 ISO-8859-2       ro       0.36
## 3 ISO-8859-9       tr       0.20
## 4   UTF-16BE                0.10
## 5   UTF-16LE                0.10

Nevertheless, encoding detection is an operation that relies on heuristics. Therefore, there is a chance that the output might be imprecise or even misleading.

Converting Between Encodings#

Knowing the desired source and destination encoding precisely, stri_encode() can be called to perform the conversion. Contrary to the built-in iconv(), which relies on different underlying libraries, the current function is portable across operating systems.

y <- stri_encode(x, from="ISO-8859-1", to="UTF-8")

stri_enc_list() provides a list of supported encodings and their aliases in many different forms. Encoding specifiers are normalised automatically, e.g., "utf8" is a synonym for "UTF-8".

Splitting the output into text lines gives:

tail(stri_split_lines1(y), 4)  # spoiler alert!
## [1] "A Homero iguala si a escrebir intenta," 
## [2] "y a tanto llega de su pluma el vuelo,"  
## [3] "cuanto es verdad que a todos es notorio"
## [4] "el alto ingenio de don DIEGO OSORIO."

Normalising Strings#

In the section on Testing String Equivalence we’ve provided some examples of canonically equivalent strings whose code point representation was different. Unicode normalisation forms C (Canonical composition, NFC) and D (Canonical decomposition, NFD) can be applied so that they’ll compare equal using bytewise matching.

x <- "a\u0328 ą"   # a, combining ogonek, space, a with ogonek
stri_enc_toutf32(  # code points as decimals
  c(x, stri_trans_nfc(x), stri_trans_nfd(x)))
## [[1]]
## [1]  97 808  32 261
## [[2]]
## [1] 261  32 261
## [[3]]
## [1]  97 808  32  97 808

Above we see some example code points before, after NFC, and after NFD normalisation, respectively.

It might be a good idea always to normalise all the strings read from external sources (files, URLs) with NFC.

Compatibility composition and decomposition normalisation forms (NFKC and NFKD, respectively) are also available if the removal of the formatting distinctions (font variants, subscripts, superscripts, etc.) is desired. For example:

## [1] "r2{"

Such text might be much easier to process or analyse statistically.