This document covers the following topics:
In Natural, you can specify Unicode strings with the format U and U constants.
Format U
With format U, you can define data which holds Unicode strings. The
Natural data format U is internally UTF-16.
See also Format and Length of User-Defined Variables in the Programming Guide.
U Constants
You can define Unicode constants with the prefix
"U". For example:
U'Äpfel'
The prefix "UH" can be used for defining Unicode constants in hexadecimal format. Four hexadecimal digits represent one UTF-16 code unit as defined by the Unicode Standard. So the overall length must be a multiple of four. For example, if you need the hexadecimal form of
U'Äpfel'
you need the UTF-16 code units for "Ä", "p", "f", "e" and "l" (which are "U+00C4", "U+0070", "U+0066", "U+0065" and "U+006C") and you have to combine them to the following hexadecimal string:
UH'00C4007000660065006C'
See also Unicode Constants in the Programming Guide.
The data format U is endian-dependant. This has to be considered when moving between the formats B and U.
The advantage of the U format (as compared with the A format) is, that
it can hold any combinations of characters from different languages and that it
does not depend on the default code page (value of the system variable
*CODEPAGE
).
Moreover, the U format makes it easier to share data between different
platforms; no more conversions (for example, from EBCDIC to ASCII) are
necessary.
On the other hand, U format data consumes more memory than A format data. For languages in which most strings can be represented by single-byte encoding, U format will result in strings occupying twice the space that was previously required. However, for East Asian languages, the memory consumption will often not be higher.
Basically, U format can be used in most statements which allow A format.
However, if a Natural object name is given as an operand of a statement (for
example, in the CALLNAT
statement), U cannot be
used because Natural object names have A format. For information on a specific
statement, see the Statements
documentation.
Basically, A and U format can be used together in one statement, for example:
EXAMINE S FOR P WITH DELIMITER D REPLACE R
where S
is U format, and
P
, D
and
R
are A format.
In the above example, the variables
P
, D
and
R
are temporarily converted into the target format
U before the actual execution of the EXAMINE
statement. The
conversion from Unicode to code page or vice versa requires calling an ICU
function. The conversion requires additional computing time and additional
memory. This disadvantage is even greater with very large variables. To avoid
frequent conversions, it is recommended that you use only one format within one
statement. When all operands in the above example are specified in either U
format or A format, a conversion is not necessary. However, if you choose to
specify only U operands, this variant will be slower since (due to its nature)
this operand type consumes more resources; one character is then coded with 2
bytes (instead of 1 byte which is used with A format).
With a conversion (especially from U format to A format), there
is always the risk that characters cannot be represented in the target code
page. For example, you want to convert the Unicode character
"U+05D0" (Hebrew letter Alef) into the code page
IBM01140 (English). Since this character is not contained in the code page
IBM01140, either the substitution character for this code page is used, or the
place holder which was specified when defining the code page in
NATCONFG
(mainframe only). When the parameter
CPCVERR
is set to ON
, an error message will be issued in this case,
indicating a conversion error. In any case, the original information will be
lost.
The following statements are particularly affected when using Unicode:
Normalization in Unicode: A process of removing alternate representations of equivalent sequences from textual data in order to convert the data into a form that can be binary-compared for equivalence. The Unicode Standard defines different normalization forms. The normalization form that results from the canonical decomposition of a Unicode string, followed by the replacement of all decomposed sequences by primary composites where possible, is called "Normalization Form Composed" (NFC).
Natural assumes that all Unicode data is in NFC format to assure that
string operations can be performed without partial truncation of a Unicode
character. Natural conversion operations assure that the resulting Unicode
string is in NFC. If Unicode data is received from outside of Natural and it is
not guaranteed that the data has NFC format, the
MOVE
NORMALIZED
statement can be applied.
Example:
Character Sequence | NFC |
---|---|
ê (U+00EA) | ê (U+00EA) |
e (U+0065) + ^ (U+0302) | ê (U+00EA) |
Note:
Concatenating two or more strings in NFC format can result in
not-NFC format.
An implicit conversion between Unicode and the default code page (value
of the system variable *CODEPAGE
)
is performed when moving strings from U to A or vice versa with the
MOVE
statement.
Furthermore, the MOVE
ENCODED
statement can be used for conversion between
different code pages or from any available code page to Unicode and vice versa.
This can be helpful if data is coming from outside of Natural and this data is
coded in a code page which differs from the default code page. But even for
conversions between the default code page and Unicode, this statement can be
used if you want to obtain a potential conversion error with the
GIVING
clause; if CPCVERR
is set to
ON
, the MOVE
statement will stop with a runtime error
in this case.
If a character cannot be converted, it depends on the setting of the
CPCVERR
parameter whether a substitution character is
used for this character or whether the conversion fails. On Windows, UNIX and
OpenVMS platforms, the default substitution character (defined by ICU) for the
conversion from Unicode to the default code page (CP
) can be changed with
the profile parameter SUBCHAR
.
This statement can also be used for conversion from U data into UTF-8 format.
Note:
If you convert data to a code page which differs from the default
code page, it is recommended not to use this data in I/O. I/O is only
meaningful with the default code page.
A "grapheme" is what a user normally thinks of as a character. In most cases, a Unicode code point is a grapheme, however, a grapheme can also consist of several Unicode code points. For example, a sequence of one base character and one or more combining characters is a grapheme.
Example: "a" (U+0061) + "." (U+0323) + "^" (U+0302) defines one grapheme which is displayed as follows:
Note:
If a base/combining character sequence is normalized, this does not
mean that the sequence is always replaced by a pre-composed character, because
not all characters are available in a pre-composed format.
A "supplementary code point" is a Unicode code point between "U+10000" and "U+10FFFF". A supplementary code point is in UTF-16, represented by a surrogate pair which consists of two code units where the first value of the pair is a "high-surrogate code unit", and the second is a "low-surrogate code unit". Such characters are generally rare, but some are used, for example, as part of Chinese and Japanese personal names, and therefore support for these characters is commonly required for government applications in East Asian countries.
The string handling statements such as
EXAMINE
and its
SUBSTRING
option work on UTF-16 code units. It is the user's
responsibility that the code does not separate graphemes or surrogate pairs.
However, the clauses CHARPOSITION
and
CHARLENGTH
of the EXAMINE
statement (see
Syntax 3 -
EXAMINE for Unicode Graphemes) can be used to ask for the
start and length (in UTF-16 code units) of graphemes. The result values can be
used for SUBSTRING
calls. With these clauses, it is possible to
scan a string grapheme by grapheme.
Example:
DEFINE DATA LOCAL 1 #UNICODE-STRING (U15) 1 #CODE-UNIT-INDEX (N4) 1 #CODE-UNIT-LEN (N4) 1 #GRAPHEME-NUMBER (N4) END-DEFINE MOVE U'' TO #UNICODE-STRING #GRAPHEME-NUMBER := 1 REPEAT EXAMINE FULL VALUE OF #UNICODE-STRING FOR CHARPOSITION #GRAPHEME-NUMBER GIVING POSITION IN #CODE-UNIT-INDEX GIVING LENGTH IN #CODE-UNIT-LEN DISPLAY #UNICODE-STRING #GRAPHEME-NUMBER #CODE-UNIT-INDEX #CODE-UNIT-LEN #GRAPHEME-NUMBER := #GRAPHEME-NUMBER + 1 WHILE #CODE-UNIT-INDEX NE 0 END-REPEAT END
The above example program provides the following output:
Page 1 05-12-15 09:33:49 #UNICODE-STRING #GRAPHEME-NUMBER #CODE-UNIT-INDEX #CODE-UNIT-LEN --------------- ---------------- ---------------- -------------- 1 1 1 2 2 2 3 4 1 4 5 3 5 8 1 6 9 3 7 12 1 8 13 3 9 0 0
XML documents can contain information within the XML document header
about the encoding of the document (for example, <?xml version="1.0"
encoding="UTF-8" ?>
). If an XML document contains this information,
the parsing of the XML document on Windows, UNIX and OpenVMS platforms always
includes a conversion of the code page given within the XML document header to
the default code page of Natural (value of the system variable
*CODEPAGE
),
if the receiving field is not of format U.
On mainframe platforms, the document to be parsed is always internally converted to UTF-16 (if the document is not already encoded in UTF-16).
See the description of the PARSE
XML
statement for further information.
See also Statements for Internet and XML Access in the Programming Guide.
Data transfer with the REQUEST DOCUMENT
statement normally
does not involve any code page conversion. If you want to have the outgoing
and/or incoming data encoded in a specific code page, you can use the
DATA ALL
clause and/or the RETURN PAGE
clause of the
REQUEST DOCUMENT
statement to specify this.
See the description of the REQUEST DOCUMENT
statement for
further information.
See also Statements for Internet and XML Access in the Programming Guide.
On mainframe platforms, the
DEFINE
PRINTER
statement provides a CODEPAGE
clause to
provide for conversion of print report data into a code page different from the
default code page (value of the system variable
*CODEPAGE
).
On Windows, UNIX and OpenVMS platforms, the
DEFINE
PRINTER
statement does not provide such a clause; if the
CODEPAGE
clause is defined, it is ignored on Windows, UNIX and
OpenVMS platforms.
Data exchange in Unicode format via RPC is supported. See the
description of the CALLNAT
statement.
If U data is sent from a platform with big endian encoding to a platform with little endian encoding or vice versa, the encoding is adapted so that it conforms with the encoding on the receiving platform. For example, when U data in little endian encoding arrives on a big endian platform, this data is converted to big endian encoding before it is handed over to the program. When this data is sent back, it is converted back to little endian encoding.
In a logical condition criterion, Unicode operands can be used together with alphanumeric and binary operands. If not all operands are Unicode operands (format U), the second and all following operands are converted to the format of the first operand. If a binary operand (format B) is specified as the second or a following operand, the length of the binary operand must be even; the binary operand is assumed to contain Unicode code points.
If the first operand is a Unicode operand (format U) and the comparison is therefore performed as a Unicode comparison, the ICU collation algorithm is used. The ICU algorithm does not perform a plain binary comparison. For example,
some code points such as "U+0000" are ignored during the comparison process,
combined characters are considered as being equal to the equivalent single code point (for example, the German character "ä" represented by "U+00E4" and the combination of the code points "U+0061" and "U+0308" are considered as being equal by ICU).
Note:
Comparing an alphanumeric and a Unicode operand can deliver different
results, depending on the sequence of the fields.
See also Logical Condition Criteria in the Programming Guide.
This section covers the following topics:
The system variable *CODEPAGE
is used to return the IANA name of the default code page, that is, the code
page used for conversions between Unicode and code page format.
The system variable *LOCALE
contains the language and country of the current locale.
U format can be used for large and dynamic variables. For dynamic U
variables, *LENGTH
returns the number of UTF-16 code units.
See also Introduction to Dynamic Variables and Fields in the Programming Guide.
The following session parameters are available:
Parameter | Description |
---|---|
DL |
Specifies the display length for a field of format A or U. See also Display Length for Output - DL Parameter in the Programming Guide. |
EMU * |
Edit mask in Unicode. |
ICU * |
Insertion character in Unicode. |
LCU * |
Leading characters in Unicode. |
TCU * |
Trailing characters in Unicode. |
Session parameters marked with an asterisk (*) in the above table are only available on Windows, UNIX and OpenVMS platforms.
DL
versus
AL
As long as Natural was not Unicode-enabled, the length of an alphanumeric field was always identical to the number of columns needed for displaying the field (called number of display columns). This was even true for the East Asian languages which use DBCS code pages: an A format field can hold only half the characters (for example, A10 results in A5).
Example:
DEFINE DATA LOCAL 1 #A8 (A8) END-DEFINE #A8 := 'computer' WRITE #A8 #A8 := '' WRITE #A8 END
The above code results in the following output:
Page 1 ... computer
With U format fields, the length of a field and the number of display
columns is no longer identical. U characters can have narrow width (for
example, Latin characters), wide width (for example, Chinese characters) or no
width (for example, combining characters). Therefore, it is totally unknown how
many display columns a U field needs; this depends on the contents of the
field. Natural cannot automatically decide how many columns are to be reserved
on the screen: if the maximum size is assumed, Latin output will have large
gaps, and if the minimum size is assumed, Chinese output cannot be displayed
totally. Therefore, the Natural programmer has to define the display width of a
field; this is done with the DL
parameter. The
AL
parameter cannot be used for this purpose, because it cuts away the part of the
field which exceeds the defined length. But we do not want to cut any
characters from the U field; we only want to define the start position of the
following field.
Example:
DEFINE DATA LOCAL 1 #U8 (U8) 1 #U4 (U4) END-DEFINE #U8 := 'computer' WRITE #U8 #U4 := U'' WRITE #U4 (DL=8) END
The above code results in the same output as above:
Page 1 ... computer
On Windows, either locally with the
output window
or in a remote development environment with the Natural Web I/O Interface
client, it is possible to scroll in a field where the defined value for the
DL
parameter is smaller than the real display width of
the field.
EMU
, ICU
,
LCU
, TCU
versus
EM
, IC
,
LC
, TC
The parameters EMU
,
ICU
,
LCU
and
TCU
(which are only available
on Windows, UNIX and OpenVMS platforms)
allow using characters which are not included in the default code page.
They are stored in Unicode format in the generated program. These parameters
can be used with all field formats.
The parameters EM
,
IC
,
LC
and
TC
can
also be used with U format fields. These parameters may also be useful if
characters which are contained in the default code page have different
encodings in other code pages. For example, the Euro sign (€) has the code
point "0x80" in the
"windows-1252" (Latin 1) code page, but the code
point "0x88" in the
"windows-1251" (Cyrillic) code page. Thus, using a
Unicode parameter will assure that the Euro sign is always displayed correctly,
no matter what code page is installed on the PC.
Example for EMU
:
DEFINE DATA LOCAL 01 EMPLOYEES-VIEW VIEW OF EMPLOYEES 02 FIRST-NAME 02 NAME 02 SALARY (1) END-DEFINE * READ (6) EMPLOYEES-VIEW DISPLAY NAME FIRST-NAME SALARY(1) (EMU=999,999) END-READ * END
The above code results in the following output:
Page 1 05-12-15 11:45:36 NAME FIRST-NAME ANNUAL SALARY -------------------- -------------------- -------- ADAM SIMONE 159,980 MORENO HUMBERTO 165,810 BLOND ALEXANDRE 172,000 MAIZIERE ELISABETH 166,900 CAOUDAL ALBERT 167,350 VERDIE BERNARD 170,100
The library SYSEXPG
contains sample programs for Unicode
and code page support in Natural:
UNICOX01
lists all Unicode characters.
UNICOX02
converts Unicode characters to code points and
vice versa.
CODEPX01
lists all code pages, whether the code page is
supported in Natural and which encoding it uses. For all supported code pages,
it offers services to list the characters of the code page and to convert a
string from the code page into its hexadecimal representation and vice
versa.
CODEPXL1
lists all characters of any 1-byte code
page.
CODEPXL2
lists all characters of any 2-byte code
page.
CODEPXC1
converts a string from any code page into its
hexadecimal representation and vice versa.