Code Converter With Truncation Processing

1. Field of the Invention

The present invention relates to a system for converting between character codes for written or displayed text, and more particularly, to a code converter for converting between one character set and another character set.

2. Description of the Related Art

Computers and other electronic devices typically use text to interact with users. The text is usually displayed on a monitor or some other type of display device. Because the text must be represented in digital form inside the computer or other electronic device, a character set encoding must be used. Generally speaking, a character set encoding operates to encode each character of the character set with a unique digital representation. The characters (which are encoded) correspond to letters, numbers and various text symbols are assigned numeric codes for use by computers or other electronic devices. The most popular character set for use with computers and other electronic devices is the American Standard Code for Information Exchange (ASCII). ASCII uses 7-bit sequences for its encodings. In other countries, different character sets are used. In Europe, the dominant character encoding standards are the ISO 8859-X family, especially ISO 8859-1 (called "Latin-1") developed by the International Standards Organization (ISO). In Japan, the dominant character encoding standard is JIS X0208 where JIS refers to the Japanese Information Standard and was developed by Japan Standards Association (JSA). Examples of other existing character sets include Mac.TM. OS Standard Roman encoding (by Apple Computer, Inc.), Shift-JIS (Japan), Big5 (Taiwan), and many more.

With the ongoing globalization of business and networks, it has become important for computers or other electronic devices to be able to handle multiple character encodings. For example, the same computer or electronic device may be used by persons of different nationalities who wish to interact with the computer or other electronic device in a different language. For each such language a different character set encoding is usually needed. However, character sets for the same language can also differ.

There is also a need to be able to convert from one character set encoding to another encoding. For example, a user in France using ISO 8859-1 may want to send an electronic mail message in French to a user in Israel who is using ISO 8859-8. Because the sender and receiver are using different character set encodings, the non-ASCII characters in the message will be garbled for the user in Israel. Ideally, one of the computers or electronic devices would convert from one character set to another character set. This has been achieved to a limited extent between a few character sets, but is largely not possible with modern computers or electronic devices. Code conversion is made difficult because of the numerous different character standards and the often conflicting or inconsistent national standards.

The Unicode.TM. standard (hereafter simply Unicode or Unicode standard) was developed to provide an international character encoding standard. The designers of the Unicode standard wanted and did provide a more efficient and flexible method of character identification. The Unicode standard includes characters of all major International Standards approved and published before Dec. 31, 1990, as well as other characters not in previous standards. The characters are encoded in the Unicode standard without duplication. The codes within the Unicode standard are 16-bits (or 2 bytes) wide.

A character code standard such as the Unicode standard facilitates code conversion and enables the implementation of useful processes operating on textual data. For example, in accordance with the above example, the computer or other electronic device in France can transmit Unicode characters and the computer or other electronic device in Israel can convert the Unicode characters it receives into a Hebrew based character set that is compatible with the computer or other electronic device in Israel.

Below is a general overview discussion of the Unicode standard. For additional detail about the Unicode standard, see, e.g., The Unicode Standard, Worldwide Character Encoding, Version 1.0, Addision-Wesley 1991 (Version 1.1 is also available for additional details), both versions of which are hereby incorporated by reference in their entirety.

The design of the Unicode encoding scheme is independent of the design of basic text processing algorithms, with the exception of directionality. Unicode implementations are assumed to contain suitable text processing and/or rendering algorithms. For convenience, all codes in the Unicode standard are grouped by a linguistic and functional category, though all the codes in the Unicode standard are equally accessible. "The code space in the Unicode standard is divided into six zones: General Scripts (alphabetic and other scripts that have relatively small character sets), Symbols, CJK (Chinese, Japanese, and Korean) Auxiliary, CJK ideographs, Private Use and Compatibility. The General Script zone covers alphabetic or syllabic scripts such as Latin, Cyrillic, Greek, Hebrew, Arabic, Devanagari and Thai. The Symbol zone includes a large variety of characters for punctuation, mathematics, chemistry, dingbats, and so on. The CJK Auxiliary zone includes punctuation, symbols, Kana, Bopomofo, and single and composite Hangul. The CJK Ideographic zone provides space for over 20,000 ideographic characters or characters from other scripts. The Private Use zone is used for defining user- or vendor- specific graphic characters. The Compatibility Zone contains characters from widely used corporate and national standards that have other canonical representations in Unicode encoding." The Unicode Standard, supra, Version 1.0, p. 13.

The Unicode standard provides character properties and control characters. Character properties are useful for use in parsing, sorting, and other algorithms requiring semantic knowledge about the code points within the Unicode encoding. The character properties identified by the Unicode standard include: digits, numbers, space characters, non-spacing marks, and direction. The Unicode characters are grouped based on the characters properties. Digits, numbers and space characters are well known. The non-spacing marks group houses non-spacing marks, and the direction group houses the direction characters.

Non-spacing marks (e.g., accent marks in Greek and Roman scripts, vowel marks in Arabic and Devanagari) do not appear linearly in the final rendered text. In a Unicode character code sequence, all such characters follow the base character which they modify, or the character after which they would be articulated in phonetic order (for example, Roman "A" is stored as "A" when not stored in a precomposed form). When rendered, these characters (i.e., non-spacing marks) are intended to be positioned relative to the preceding base character in some manner, and not themselves occupy a spacing position.

Control characters are encoded in the Unicode standard, but are not themselves graphic characters. These control characters can, for example, be used for indicating a horizontal tab, or supplying additional information about text such as formatting attributes or structure, or for controlling by directional formatting. Because the Unicode standard also provides bidirectional character ordering, the Unicode encoding scheme also includes characters to specify changes in direction. For example, Greek, Roman and Thai have a dominant direction of left-to-right, while Arabic and Hebrew have a dominant direction of right-to-left. The use of control characters to change directions will sometimes be needed by the user or system, for example, when left-to-right characters are to be displayed in right-to-left order.

When receiving text over a network, the data associated with the text typically arrives in blocks of data. The data is completely received only after all the blocks comprising the data have been received. The received data is placed in a buffer where it awaits conversion of character codes. The buffer, however, in many cases is unable to store the entire data stream or sometimes even a block thereof. In any case, the end of the buffer (i.e., the last character in the buffer) may occur in the middle of what is later to be determined by the scanner 408 to be a text element. When the end of the buffer occurs in the middle of a text element, the scanner 408 is unable to correctly obtain the last text element because subsequent characters may effect or be part of the last text element.

Thus, there is a need for a code converter that is able to accurately, as well as flexibly, convert characters from a source character set to characters of a target character set while taking into consideration the context of the characters.

Broadly speaking, the invention is a truncation processing technique that ensures that a source string received for conversion is accurately converted into a different, target character encoding even when the source string exceeds the length of an input buffer which holds the source string for conversion.

The truncation processing operates to truncate a portion of the source string held in the input buffer so that the truncated portion is able to be converted to the target character encoding without being affected by subsequent characters in the source string. Of course, the portion of the source string held in the input buffer after truncation (truncated portion) should be maximized so that the code conversion can be carried out efficiently. The invention is particularly useful when the input source string is data received over a network. For example, the data may be textual data being electronically transferred over the Internet. In any case, the invention can be implemented in numerous ways, including as a method, apparatus or system, or on a computer readable medium.

As a code conversion system for converting a source string to a target string, an embodiment of the invention includes: a converter for controlling the conversion of the source string having a first character encoding into the target string having a second character encoding; a buffer for receiving a portion of the source string at a time, the source string including more than one portion; a truncator for truncating the portion of the source string; a scanner for dividing the truncated portion of the source string into text elements, each text element including one or more characters of the truncated portion of the source string; a mapping table for storing target encodings for text elements of the source encoding; and a lookup handler for looking up in the mapping table a conversion code associated with a second character encoding for each of the text elements.

As a first method for truncating a source string for character conversion to a target string, an embodiment of the invention performs the operations of: receiving a buffer portion of a source string in a buffer, the source string including more than one buffer portion; determining a subpart of the buffer portion of the source string that can be converted to a target string without being affected by subsequent buffer portions of the source string; and converting the subpart of the buffer portion of the source string into target encodings. Preferably, the first method also performs the operation of saving a remaining part of the buffer portion for conversion with a next buffer portion or subpart thereof, where the remaining part together with the subpart equals the buffer portion.

As a second method for truncating a source string for character conversion to a target string, an embodiment of the invention performs the operations of: receiving a portion of a source string in a buffer; determining a text element within the portion of the source string, each text element including one or more characters of the source string; determining whether the text element is complete; including the text element in a truncated portion of the source string when the text element is complete; repeating the text element determining until the portion of the source string has been completely considered; and thereafter outputting the truncated portion of the source string for character conversion. Preferably, the second method also performs the operation of saving any remaining portion of the source string for use with a next portion of the source string received in the buffer.

As a computer readable medium containing program instructions for convening a source string into a target string, an embodiment of the invention includes: computer readable code configured to cause a computer to effect receiving a portion of a source string having a first character encoding in a buffer; computer readable code configured to cause a computer to effect truncating the portion of the source string; computer readable code configured to cause a computer to effect dividing the truncated portion of the source string into text elements, each text element including one or more characters of the truncated portion of the source string; computer readable code configured to cause a computer to effect looking up in a conversion code associated with a second character encoding for each of the text elements; and computer readable code configured to cause a computer to effect combining the conversion codes for the text elements so as to form a target string of the second character encoding.

Other aspects and advantages of the invention will become apparent from the following detailed description, taken in conjunction with the accompanying drawings, illustrating by way of example the principles of the invention.

The present invention will be readily understood by the following detailed description in conjunction with the accompanying drawings, wherein like reference numerals designate like structural elements, and in which:

FIG. 1 is block diagram of a representative computer system in accordance with the invention;

FIG. 2 illustrates the format of Unicode character encodings;

FIG. 3 illustrates a block diagram of a basic Unicode code conversion system according to the invention which receives a source string and outputs a target string;

FIG. 4 illustrates a block diagram of an embodiment of a From-Unicode code conversion system according to an embodiment of the invention;

FIG. 5 is a schematic diagram of a preferred arrangement for a mapping table of the Unicode code conversion system;

FIG. 6A is a flow chart of processing performed by an application program making use of the Unicode code conversion system according to an embodiment of the invention;

FIG. 6B is a flow chart of truncation processing in accordance with an embodiment the invention;

FIG. 7 is a flow chart of Unicode converter processing according to an embodiment of the invention;

FIG. 8 is a flow chart of update offsets processing according to an embodiment of the invention;

FIGS. 9A and 9B are flow charts of next text element processing according to an embodiment of the invention;

FIG. 10 is a block diagram of a scanner according to an embodiment of the invention;

FIG. 11 is a schematic diagram of a preferred format for the attributes table shown in FIG. 10.

FIG. 12 is a flow chart of attributes lookup processing according to an embodiment of the invention;

FIGS. 13A and 13B are schematic diagrams associated with a scanner table utilized by the scanner to determine the next action according to an embodiment of the invention;

FIG. 14 is a table which represents both a preferred layout and the information which would be stored in the scanner table according to a preferred embodiment of the invention;

FIG. 15 is a flow chart of lookup text element processing according to an embodiment of the invention;

FIG. 16 is a flow chart illustrating search variant list processing according to an embodiment of the invention;

FIGS. 17A and 17B are schematic diagrams of tables and data formats associated with the variant list processing according to an embodiment of the invention;

FIG. 18 is a flow chart illustrating fallback handling processing in accordance with an embodiment of the invention;

FIG. 19 is a flow chart illustrating default processing according to an embodiment of the invention; and

FIG. 20 illustrates a block diagram of an embodiment of a To-Unicode code conversion system according to an embodiment of the invention.

Embodiments of the invention are discussed below with reference to the drawings. However, those skilled in the art will readily appreciate that the detailed description given herein with respect to these figures is for explanatory purposes as the invention extends beyond these limited embodiments.

The invention is a truncation processing technique that ensures that a source string received for conversion is accurately converted into a different, target character encoding even when the source string exceeds the length of an input buffer which holds the source string for conversion. The truncation processing technique operates to truncate a portion of the source string held in the input buffer so that the truncated portion is able to be converted to the target character encoding without being affected by subsequent characters in the source string. Of course, the portion of the source string held in the input buffer after truncation (truncated portion) should be maximized so that the code conversion can be carried out efficiently.

The truncation processing technique according to the invention is described in detail below with reference to a code conversion system that converts source characters to target characters of a different encoding.

FIG. 1 is block diagram of a representative computer system 100 in accordance with the present invention. The computer system 100 includes a central processing unit (CPU) 102. The CPU 102 is coupled bidirectionally with random access memory (RAM) 104 and unidirectionally with read only memory (ROM) 106. Typically RAM 104 includes programming instructions and data associated with code conversion, including tables as described herein, as well as other data and instructions for processes currently operating on CPU 102. The ROM 106 typically includes basic operating instructions, data and objects used by the computer system 100 to perform its functions. In addition, a mass storage device 108, such as a hard disk, CD ROM, magneto-optical (floptical) drive, tape drive or the like, is coupled bidirectionally with CPU 102. Mass storage device 108 generally includes additional programming instructions, data and text objects that typically are not in active use by the CPU 102, although the address space may be accessed by the CPU 102, e.g., for virtual memory or the like. Each of the above described computers further includes an input/output source 110 that typically includes input media such as a keyboard, pointer device (e.g., a mouse or stylus) and the like. The computer system 100 can also include a network connection 112 over which data and instructions can be transferred to a network. Additional mass storage devices (not shown) may also be connected to CPU 102 through network connection 112. The computer system 100 further includes a display screen 114 for viewing text and images generated or displayed by the computer system 100.

The CPU 102 together with an operating system (not shown) operate to execute computer code. The computer code may reside on the RAM 104, the ROM 106, or a mass storage device 108. The computer code could also reside on a portable program medium 116 and then be loaded or installed onto the computer system 100 when needed. The portable program medium 116 may, for example, CD-ROM, PC Card device, RAM device, floppy disk, or magnetic tape.

I. Definitions

1. Code Point: A code point is a bit pattern in a particular encoding. Usually the bit pattern is one or more bytes long. A Unicode code point is always 16 bits or two bytes.

2. Encoding: An encoding is a one-to-one mapping between a set of characters and a set of code points. For example, the ASCII encoding maps a set including a-z, A-Z, and 0-9 to the code points x00 through x7F.

3. Text Element: A text element is a sequence of one or more code points that are treated as a unit for a particular operation. For example, LATIN CAPITAL LETTER U followed by NON-SPACING DIAERESIS is a text element (e.g., two adjacent characters in this example) for this for the code conversion operation in accordance with the invention.

4. Glyphs: A displayed form that provides the visual representation of a character. For example, an italic "a" and a roman "a" are two different glyphs representing the same underlying character.

5. Presentation Form: A presentation form is a glyph that varies its visual form depending on the context. Some encodings map only abstract characters, which are independent of context, while other encodings map only presentation forms. For example, a ligature such as "fi" is a presentation form for the character sequence LATIN CAPITAL LETTER F followed by LATIN CAPITAL LETTER I.

6. Fallback: A fallback is a sequence of one or more code points in the target encoding that are not exactly equivalent to the source code points but which preserve some of the information of the original. For example, (C) is a possible fallback for .COPYRGT..

7. Default: A default is a sequence of one or more code points in the target encoding that are used when nothing in the target encoding even resembles the source code points.

II. Unicode Converter

The conversion technique according to the invention converts source characters to target characters of a different encoding. Preferably, either the source characters or the target characters are Unicode characters.

The Unicode standard is a compilation of character encodings developed into a single, universal, international character encoding standard. FIG. 2 illustrates the format of the Unicode character encodings. Specifically, the Unicode standard provides for codes which are 16 bits wide as illustrated by a format 200 shown in FIG. 2. Within this document, Unicode characters are represented in hexadecimal with a preceding u (e.g., u0041), and characters in other encodings are represented in hexadecimal with a preceding x (e.g., x41 for a 1-byte character, x8140 for a 2-byte character).

FIG. 3 illustrates a block diagram of a basic Unicode code conversion system 300 according to the invention which receives a source string 302 and outputs a target string 304. The Unicode code conversion system 300 operates to convert characters of the source string 302 into one or more characters in the target stream which are of a different character encoding than the encoding utilized in the source string. Preferably, the Unicode code conversion system 300 converts from Unicode to a different target encoding (From-Unicode) or converts from a different source encoding (To-Unicode).

FIG. 4 illustrates a block diagram of an embodiment of a Unicode code conversion system 400 according to the invention. A Unicode code conversion system 400 includes a From-Unicode converter 402 which receives a Unicode string 404 and produces a target string 406. The Unicode string is stored in a buffer 405. A truncator 407 is provided to truncate the length of the Unicode string 404 stored in the buffer 405 so that accurate conversion is ensured.

The From-Unicode converter 402 controls the overall operation code conversion. In so doing, the From-Unicode converter 402 interacts with a scanner 408. The scanner 408 in conjunction with a scanner table 410 scans the truncated Unicode string 404 (supplied by the truncator 407) to identify a text element. The From-Unicode converter 402 then uses a lookup handler 412 to look up the one or more characters in target encoding for the text element identified by the scanner 408. The lookup handler 412 uses a mapping table 414 to obtain the one or more characters in the target encoding for the text element. Additionally, the From-Unicode converter 402 may also use a fallback handler 416. The fallback handler 416 operates together with the mapping table 414 to identify one or more characters in the target encoding that are able to be used as a fallback mapping for the text element in cases where the look-up handler 412 has been unable to identify one or more characters in the target encoding for the text element. A state administrator 418 maintains or stores information on the current state of the conversion. This information may, for example, include context, direction and state of symmetric swapping. Such information is needed for non-block delimited conversion in which the truncated Unicode string does not end at the end of a block. In such a case, by storing information on the current state of the conversion, the code conversion process provides correct code conversions even when the input string 404 is larger than the size of the buffer 405.

A. Scanner and Scanner Table

The scanner 408 in conjunction with the scanner table 410 scans the truncated Unicode string 404 and returns the next text element and any additional information needed by the look-up handler 412. The additional information includes one or more of direction information, context information, and various state indicators. The general operation of the scanner 408 is as follows. The scanner 408 scans through the characters of the truncated Unicode string 404. If direction information is needed for the target encoding, then the character direction is obtained for each character in the text element. Also, if character form context information is needed for the target encoding, then character form context information is obtained for each character in the text element. Then, as the scanner 408 scans through each of the characters, the scanner 408 takes an action for the character in accordance with information residing in the scanner table 410. The particular action that the scanner 408 takes is determined based on state and character class. The actions that the scanner 408 can take include: marking the current character, setting or clearing the symmetric swapping bit, noting the contextual form of a text element, setting a flag that indicates that the text element will need reordering, and indicating end of the text element. The symmetric swapping bit, the context and the direction are saved by the state administrator 418 as information pertaining to the state of the scanner. Before returning, the scanner 408 saves context information for the text element. The scanner 408 returns the text element (each text element within the input string) and its attributes. The attributes include the following: direction, class, priority, symmetric swapping state, subset and context. After the scanner 408 determines a text element, then the characters may need to be reordered into canonical order. As an example, reordering of the characters within a text element is done when the text element includes non-spacing marks that are not in canonical order as defined by Unicode.

Preferably, the scanner 408 together with the scanner table 410 are implemented as a pair of state machines that operate in parallel. A first state machine resolves the character direction, and a second state machine computes text elements and character form context information where applicable and also maintains the symmetric swapping state. By using two separate state machines, the Unicode code conversion system 400 is easier to design and maintain. The first and second state machines can be implemented as two-dimensional arrays (or tables) indexed by state and class. In cases where the action the scanner 408 is to take depends on the character direction, then the state machine entry is an index into another table which contains the appropriate action for the scanner 408 to take for each direction.

The function of the scanner 408 is to convert the truncated Unicode string 404 into text elements and to return the text elements and their attributes. The scanner 408 needs to save certain characteristics of the text element so that it can be properly converted in the target encoding. Namely, the characteristics include the direction, the context and the symmetric swapping state. However, the scanner 408 need not know what the target encoding is because its operation is independent of the particular target encoding. Nevertheless, the Unicode code conversion system 400 is preferably implemented such that the definition of a text element (i.e., the chunking behavior) could vary with the target encoding simply by modifying the scanner table 410.

The directionality of characters is used for presentation of the characters. For example, when Arabic or Hebrew are displayed on a display screen, they are ordered from right-to-left. Most Unicode characters have an implicit direction, see Unicode Version 1.0, at p. 407 (Section 4.6) and p. 611 (Appendix A). The implicit direction classes provided with the Unicode standard and their values include: Left-Right (0), Right-Left (1), European Number (2), European Number Separator (3), European Number Terminator (4), Arabic Number (5), Common Number Separator (6), Block Separator (7), Segment Separator (8), Whitespace (9), and Other Neutrals (10). The scanner 408 looks up the direction class for characters of the text element. The direction class is then used to resolve the direction of the text element. There are also special Unicode characters which cause overriding or embedding of directionality. These special direction Unicode characters are treated by the scanner 408 as single character text elements.

There are some basic rules that the scanner 408 follows in forming the text elements. The base rule is that if none of the rules apply, then the text element is a single Unicode character. Another rule is that non-spacing or combining marks following a base character are grouped with the base character as a single text element. Yet another rule is that characters associated with symbols (e.g., Korean Hangul Jamos characters), ligatures or ideographs are encountered, they are combined into text elements. Still another rule is that when a fraction slash is surrounded on each side by a sequence of one or more decimal digits, they are combined as a numeric fraction text element.

The rule for non-spacing or combining marks is now explained in greater detail. According to the Unicode standard, non-spacing marks follow the base character. Hence, the non-spacing marks that follow a base character become part of the text element that includes the base character. See, The Unicode Standard, Version 1.0, at p. 403 (Section 4.5). For example, when a single, non-spacing character is followed by a character that is not a non-spacing character, then the non-spacing a character is combined with the previous character as a text element. The length of the text element is then two and the attributes for the text element are defined by the base character. If there is no preceding character, the non-spacing character is just passed on as a single text element. Multiple non-spacing characters can also be combined in this manner.

The rule for Korean Hangul Jamos characters is now explained in greater detail. Each Hangul character has an implicit value which is of one of the classes: Choseong (initial), Jungseong (medial) or Jongseong (final). The Unicode Standard, Version 1.1, (Section 5), lists the codes and the permitted combinations for these characters. The scanner 408 will group the Korean Jamos characters according to the permitted combinations of these characters. For input combinations which are not permitted, the scanner 408 will return the character as a single text element. As before, when a Hangul syllable is followed by a combining mark, the combining mark is included within the text element for the Hangul syllable.

The rule for numeric fractions is now explained in greater detail. The scanner 408 initially treats each character of fraction slash numbers as if they were single character text elements. However, when a complete fraction slash sequence is encountered, the scanner 408 will join the characters associated with the sequence into a single text element. If a digit is encountered with a combining mark, the digit and the combining mark cannot be part of a fraction slash, but the digit and the combining mark can together form a text element.

Except for non-spacing characters, all Arabic characters are passed through the scanner 408 as single text elements. The Arabic form shaping state characters are also passed through the scanner as single text elements. The directional formatting codes are passed through the scanner 408 as single text elements.

B. Lookup Handler, Mapping Tables and Fallback Handler

The mapping table 414 is used by the lookup handler 412 to match an input sequence of one or more Unicode characters to an output sequence of one or more characters in the target encoding. In addition to the Unicode sequence (i.e., text element) itself, certain additional pieces of information about the input sequence are available (e.g., direction, context, symmetric swapping state, vertical forms request, fallback request, tolerance, variant), and some tables make use of this information. Preferably, the mapping table 414 also stores data needed by the fallback handler 416, though a separate table could be provided for use by the fallback handler 416.

FIG. 5 is a schematic diagram of a preferred arrangement for the mapping table 414 of the Unicode code conversion system 400. The mapping table 414 preferably includes a header portion 500 and then segments of data within the mapping table 414 partitioned based on the number of characters in the text element. Namely, the mapping table 414 illustrated in FIG. 5 supports encodings for text elements of one to N characters. When the look-up handler 412 searches the mapping table 414 for target encodings for one-character text elements, the segment 502 of the mapping table 414 would be utilized. Similarly, if the text element is two-characters wide, then segment 504 would be used, and if the text element was N characters wide, then segment 506 would be used. Although FIGS. 4 and 5 illustrate a single mapping table 414, the Unicode code conversion system 400 uses a plurality of different mapping tables 414, namely one mapping table for each target character set. Each mapping table includes multiple subtables.

The mapping tables 414 are designed with size and overall conversion speed requirements kept in mind. The mapping tables 414 should be as small as possible without severely degrading lookup time, and the lookup time should be as fast as possible without significantly increasing the table size. The Unicode code conversion system 400 supports multiple table formats so that different formats are possible for each subtable, thereby enabling the speed/size tradeoff to be adjusted for a particular table as needed. Preferably, the design of the mapping table 414 should make the mapping from a single Unicode character to a single character in the target encoding as fast as possible since it is the most common case.

The mapping table 414 design is such that the tables support at least some of the needs of the fallback handler 416, support multiple mapping tolerances, and support multiple target character set variants. The table format is also capable of mapping Unicode sequences of one or more characters to an output sequence of zero or more characters. The mapping tables 414 can also specify several possible output sequences for a single input sequence, with the particular output sequence determined by attributes such as direction, context and symmetric swapping state. The mapping tables are also easily extended, so as to facilitate customization of the encoding behavior of the Unicode code conversion system 400.

The information needed by the mapping tables 414 includes the following: the text element from the scanner 408 (i.e., input sequence of characters to convert, with any combining marks in canonical order); whether vertical forms should be used instead of horizontal forms where available; the resolved direction of the text element; the context information for the text element (initial, medial, final, or isolated); current state of symmetric swapping (activated or inhibited); information about what level of lookup to be invoked (i.e., tolerance level (strict or loose) and fallbacks (on or off)); and an identifier for a particular encoding variant (a variant identifier). The information about the lookup level is provided by a call or an application program which calls the Unicode code conversion system 400. In a language or character set in which direction and context are unimportant, the resolved direction and the context information are not needed.

The definition of variants, the actual mappings from Unicode sequences to target sequences, and the table formats used to access them are alterable by the design of the mapping table 414. Hence, the accuracy and to some extent the performance/size tradeoffs are highly dependent the design of the mapping tables 414. It is preferable that the mapping tables 414 support strict and loose mappings, fallback mappings, and default mappings.

Strict mappings are code conversions in which round trip fidelity is guaranteed. Strict mappings from Unicode to a target character set are the converse of the mappings from that character set to Unicode. Hence, round-trip mapping from Unicode to some other character set and back again is possible when using only Unicode characters for which strict mappings to the other character set exist. Loose mappings from Unicode to the target character set are additional mappings that fall within the range of definition or established usage for the characters in the target character set. Loose mappings appear properly mapped but have some ambiguity. For example, in many character sets, a single character may have multiple semantics, either by explicit definition, ambiguous definition, or established usage. For example, the Shift-JIS character x8161 is specified to have two meanings: "double vertical line" and "parallel". Each of these meanings corresponds to a different Unicode character: u2016 "DOUBLE VERTICAL LINE" and u2225 "PARALLEL TO". When mapping from Shift-JIS to Unicode, the code conversion system must choose one of these Unicode characters, say DOUBLE VERTICAL LINE. When mapping from Unicode to Shift-JIS, the code conversion system can--and normally should--map both Unicode characters to the same Shift-JIS character. However, only one of these From-Unicode mappings is the converse of the To-Unicode mapping.

Strict verses Loose Mapping Examples:

If Unicode u000D is strictly mapped to ASCII x0D "carriage return", then Unicode u2029 "PARAGRAPH SEPARATOR" can be loosely mapped to ASCII x0D.

If Unicode u002D "HYPHEN-MINUS" is strictly mapped to ASCII x2D "hyphen-minus", then Unicodes u2010 "HYPHEN" and u2212 "MINUS SIGN" can be loosely mapped to ASCII x2D.

If Unicode u00EO "LATIN SMALL LETTER A WITH GRAVE" is strictly mapped to ISO 8859-1 xEO "small letter a with gave accent", then the two-character Unicode sequence u0061+u0300 "LATIN SMALL LETTER A" +"COMBINING GRAVE ACCENT" can be loosely mapped to ISO 8859-1 xEO.

Since Shift-JIS distinguishes halfwidth and fullwidth characters, loose mappings for Shift-JIS must also keep these distinct. Say Unicode uFF40 "FULLWIDTH GRAVE ACCENT" is strictly mapped to Shift-JIS x814D "gave accent ›fullwidth!", which is distinct from Shift-JIS x60, "grave accent ›halfwidth!". The Unicode sequence u3000+u0300 "IDEOGRAPHIC SPACE"+"COMBINING GRAVE ACCENT" can be loosely mapped to Shift-JIS x814D. However, the Unicode sequence u0020+u0300 "SPACE"+"COMBINING GRAVE ACCENT" should not be loosely mapped to Shift-JIS x814D; instead, it should be loosely mapped to Shift-JIS x60.

Fallback mappings are mappings from Unicode that do not preserve the meaning or identity of the Unicode character. That is, they map a Unicode character (or sequence of characters) to a character (or sequence) in the target set whose definition or usage does not include the meaning or usage of the Unicode character. Nevertheless, the fallback mapping, when available, provides a character (or sequence) in the target encoding that corresponds closest to the Unicode character (or sequence of characters).

Fallback Mapping Examples:

The Unicode character u0300 "COMBINING GRAVE ACCENT" can be mapped to ASCII x60 "grave accent ›spacing!" as a fallback mapping. The difference is that the Unicode character is a combining mark (non-spacing), whereas the ASCII character is a spacing mark.

The Unicode character u01C0 "LATIN LETTER DENTAL CLICK" could be mapped to ASCII x7C "vertical line" as a fallback mapping.

The Unicode character u2001 "EM QUAD" could be mapped to ASCII x20 "space" as a fallback mapping.

Hence, as illustrated in the above examples, fallback mappings are used to generate a target character (or sequence) that is a graphic approximation of the Unicode character (or sequence).

For performance reasons (i.e., the speed with which the encoding can be obtained from the mapping table), there are several possible formats for indexing into the mapping table 414. The possible formats may include a segment format, a list format, a range format, or a chain format. Separate indexes are preferably provided for different length character sequences. As a result, the index associated with each of the segments 502, 504, 506 of the mapping table 414 can be in a different format, and information at the beginning of each index specifies its format. Regardless of format, each index ultimately maps an input sequence either directly to an output sequence, or if the output sequence is long, to an offset that specifies the location of the corresponding output sequence.

The chain format for indexing into the mapping table 414 warrants further discussion. With the chain format, the beginning of the section is checked to determine whether it is a chain header for a chain format table or some other format. The chain format specifies a chain of multiple index tables, each possibly in a different format. If the desired mapping was not found in the first index table, the lookup handler 412 will then check the second, and so on. The chain format is useful, for example, when one index format (which is efficient in space and/or time) can map most but not all of the input sequences, while another less efficient index format can handle the few remaining sequences. Without a chain mechanism, the less-efficient format would have to be used for all of the index sequences. The chain format is also useful when different variants and different tolerance levels require different subtables. Each subtable in the chain has bit flags that can cause it to be excluded or included based on the mapping tolerance and variant currently being handled. When the lookup handler 412 searches the mapping table 414 for target encodings only the included subtables are considered.

These bit flags associated with each subtable form a subtable mask. Also, the caller requests (e.g., encoding variant and tolerance) and determined attributes (e.g., resolved direction and context) form a selection mask. The bit assignment within the subtable mask and the selection mask are identical. Hence, the determination of whether or not to include a particular subtable is implemented as a bitwise AND of the subtable mask with the selection mask, and then a comparison of the result with the subtable mask. If the result of the bitwise AND is the same as the subtable mask for the subtable, the subtable is included; otherwise, it is not included.

The header 500 of the mapping table 414 preferably contains:

General identifying information--format, length, checksum, and version.

The minimum target character size (in bytes) (also referred to as charsize).

General flags (for example, whether this lookup table requires direction or context data).

The maximum input sequence length handled by this table, and a list of offset/length pairs specifying the tables that handle input sequence lengths from one through this maximum.

The default fallback character or character sequence for this From-Unicode mapping.

A count and list of variants supported by this table. For each variant, one or more associated bit masks are specified. When there are multiple bit masks for a single variant, attribute information (such as direction, context, and vertical forms request) is used to determine which bit mask is used. Bits set to "1" in a bit mask are used to turn on various subtables to support the different variants.

An additional set of bit masks associated with each of the four possible tolerance settings (strict/loose, fallbacks on/off). The appropriate tolerance level mask is ORed with the variant mask to form the bit mask used to enable or disable subtables.

C. Code Conversion Processing

The processing performed by a preferred embodiment of the Unicode code conversion system 400 is explained in detail below.

FIG. 6A is a flow chart of processing 600 performed an application (i.e., calling application process or program) making use of the Unicode code conversion system 400. In particular, FIG. 6A concerns From-Unicode processing, but it should be recognized that analogous operations would be performed to convert in the other direction (To-Unicode processing). The From-Unicode converter 402 controls the overall conversion process.

Initially, the processing 600 creates and initializes 602 a new instance of state and control information for a conversion. Because the processing 600 establishes instances, multiple scanning operations can be ongoing and distinguishable by their instance. Next, a decision 604 is made based on whether truncation is needed. In the case where truncation is needed, the truncator 407 is invoked 606. Truncation is used when the input data stream exceeds the capacity of the receiving buffer 405 which holds the data for conversion. In a case where truncation is not needed or following block 606 in the case where truncation is needed, the From-Unicode converter 402 is called 608 to convert the Unicode string 404. The From-Unicode converter 402 function operates to obtain the text element and lookup the target mapping therefor as is discussed in detail below. Once the conversion function returns, the processing 600 receives the target string 406 from the From-Unicode converter 402.

The processing 600 then determines 612 whether there has been a conversion failure. If there has been a conversion failure, then the error is handled 614. On the other hand, if the conversion was successful, then a decision 616 is made based on whether the conversion is complete. The conversion is complete when the characters of the Unicode string 404 have been converted to target encodings. If the conversion is complete, then the processing 600 is complete and the target string 406 is made available to process or application which requested the code conversion. In addition, the processing 600 disposes 618 of the instance of state and control information for a conversion. On the other hand, if decision 616 determines that the conversion is not yet complete, then the processing 600 repeats blocks 604-616 until the conversion is complete or an error results.

FIG. 6B is a flow chart of truncation processing 620 in accordance with an embodiment the invention. The truncation processing 620 is invoked by block 606 in FIG. 6A and performed by the truncator 407.

The truncation processing 620 initializes 622 an output length to zero (0). The output length corresponds to the effective length of the buffer, that is, the truncated length of the buffer. Next, the next text element is obtained 624. The processing associated with obtaining 624 the next text element is performed by the scanner 408 and is described in detail below with reference to FIGS. 9A-9C. A decision 626 is then made based on whether the text element could extend beyond the physical length of the buffer (buffer length). If the text element could not possibly extend beyond the buffer length, then the output length is updated 628 to include this text element. Here, the effective length is grown text element-by-text element so long as the text element cannot possibly extend beyond the buffer length. Following block 628, a decision 630 is made based on whether additional text remains to be considered in the buffer. When there is additional text in the buffer to be considered, then processing loops back to block 624 to repeat the truncation processing 620. On the other hand, when there is no additional text in the buffer to be considered, or when block 626 determines that the text element could possibly extend beyond the buffer length, the output length is returned 632. The output length returned 632 is the effective (i.e., truncated) length of the buffer so that the buffer effectively ends at the end of a text element (namely, the last text element for the text within the buffer).

Thus, the truncation processing determines the effective (i.e., truncated) length of the buffer. This is also referred to as a truncated portion of the buffer. The extra text remaining in the buffer after truncation is referred to as a remaining portion. The remaining portion is carried over to and considered with a next buffer portion of the input source string. This processing ensures that text elements will be correctly determined.

An example of usage of the truncation processing is as follows: Example: " . . . ABCD EFG . . . "

If the portion held in the buffer ends just after "D", then the truncation processing will truncate the portion held in the buffer such that the truncated length ends after "C". It is necessary to truncate the portion in the buffer because failure to do so causes the text element "D" to be disconnected from its combining mark " " which follows. When disconnected, the text will not be properly converted to the target encoding. The remaining portion is then "D " and is carried over into the next portion, i.e., "D EFG . . . ".

FIG. 7 is a flow chart of Unicode converter processing 700. The Unicode converter processing 700 is associated with the operations performed by block 608 in FIG. 6A.

The Unicode converter processing 700 begins with a decision 702. The decision 702 determines whether there is text to be converted. If there is no text to be converted, the Unicode converter processing 700 simply returns (or completes). On the other hand, if there is text to be converted (i.e., the Unicode string 404 has not been completely processed), then processing 700 proceeds. First, offsets are updated 704 for an offset array. The offset array is an array of offsets (pointers) associated with an input string that indicate where certain changes such as font changes, line breaks, language changes, etc., occur within the input string 404 that the calling application deems significant. The updating 704 of the offset array involves adjusting the offsets (pointers) for different length characters. For example, the Unicode characters of the input Unicode string 404 are two-bytes long, while the size of the characters associated with a target encoding of ASCII is one-byte long. Here, the updating 704 of the offset array would adjust the offsets so that they point to the corresponding character in the target encoding. In effect, the offsets in the input string are mapped to the output string which has a different encoding. The next text element is then obtained 706. The scanner 408 using the scanner table 410 determines the text elements from the Unicode string 404. The obtaining 706 of the next text element is discussed in detail below. The text element obtained 706 is then looked-up 708 in the mapping table 414 to obtain a conversion code for the text element in the target encoding. The lookup is performed by the lookup handler 412 in conjunction with the mapping table 414. The looking-up 708 of the conversion code is also discussed in detail below.

Next, a decision 710 is made based on whether a conversion code was found for the text element. When a conversion code is found, an input position pointer and an output position pointer for the Unicode string 404 and the target string 406 are, respectfully, updated 712. The input position pointer indicates how much of the input string 404 has been converted. The output position pointer indicates the length of the target string 406. Following block 712, processing 700 returns to the beginning of the Unicode converter processing 700 so that the next text element (if any) of the Unicode string 404 being converted may be processed.

However, in the case when decision 710 determines that no conversion code is found in the mapping table 414, then a decision 714 is made based on whether the caller has requested fallback handling. If the caller has requested fallback handling, then fallback handling is performed 716. The fallback handling is performed by the fallback handler 416 and is discussed in detail below. On the other hand, if the caller has not requested fallback handling, then an error is indicated 718 because the text element could not be converted to the target encoding by the lookup handler 412. Following blocks 716 and 718, the input position pointer and the output position pointer for the Unicode string 404 and the target string 406 are, respectively, updated 702, and then processing returns to the beginning of the Unicode converter processing 700 so that the next text element (if any) of the Unicode string 404 being converted may be processed.

FIG. 8 is a flow chart of update offsets processing 800. The update offsets processing 800 is associated with the block 704 in FIG. 7 in which the offset array is updated.

The update offsets processing 800 begins with a decision 802 based on whether the current input position is in the offset array. If the current input position is in the offset array, then the contents of the offset array are updated 804 in accordance with the current output position length following block 804, the update offsets processing 800 returns. On the other hand, when the current input position is not within the offset array, then decision 802 simply causes the update offsets processing 800 to return because there is no offset to update.

FIGS. 9A and 9B are flow charts of next text element processing 900. The next text element processing 900 details the operations performed by the block 706 in FIG. 7 in obtaining the next text element. Preferably, the next text element processing 900 is performed by the scanner 408 in conjunction with the scanner table 410.

The next text element processing 900 begins by initializing 902 state and reorder flags. Next, a decision 904 is made based on whether the mapping table 414 needs direction information. If the mapping table requires direction information, then the direction of the next character of the Unicode string 404 is resolved 906. The processing associated with resolving 906 the direction is discussed in detail in commonly-assigned U.S. patent application Ser. No. 08/527,827 entitled "BIDIRECTIONAL CODE CONVERTER", which is hereby incorporated by reference. After the direction is resolved 906 or after the decision 904 in the case where the direction is not required, a decision 908 is made based on the context of the character. If the context of the characters within the Unicode string 404 can effect the code conversion (mapping), the context is resolved 910. The processing associated with resolving 910 the context is discussed in detail in commonly-assigned U.S. patent application Ser. No. 08/527,837, entitled "CONTEXT-BASED CODE CONVERTER", which is hereby incorporated by reference. Following block 910 or following decision 908 when context is not important, the Unicode character is obtained 912 from the Unicode string 404. Here, the next character in the Unicode string 404 is obtained. Next, the attributes for the Unicode character obtained are looked-up 914. The attributes looked-up 914 are discussed in detail below with reference to FIGS. 10-12. Then, the action for the next text element processing 900 is determined 916. The determination 916 of the action is explained in greater detail below with reference to FIGS. 13A, 13B and 14.

Next, a decision 918 is made based on whether the action is "END". If the action is not "END", then the action is at least an "ADVANCE". When the action is "ADVANCE", a decision 920 is then made based on whether the action is also "MARK". When the action is also "MARK", then the character is included 922 in the text element. Additionally, the bidirectional state is saved 924 when a "MARK" action is taken. The bidirectional state includes the directional embedding stack and the current state of the bidirectional state machine. Following block 924 or following decision block 920 when the action is not also "MARK", a switch operation 926 is performed based on an action modifier. The action modifier is a part of the action and includes modifiers such as "S", "ISS", "ASS". The action may also not have any action modifier. When the action modifier is "S", then a reorder flag is set 928. The reorder flag indicates (when set) that the characters within the text element may need to be reordered. When the action modifier is "ISS", (i.e., Inhibit Symmetric Swapping) then a swap flag is set off 930. When the action modifier is "ASS", (i.e., Activate Symmetric Swapping) the swap flag is set on 932. The swap flag indicates whether or not symmetric swapping is needed. The switch 926 can easily be adapted to include additional action modifiers using an expansion area 934. The expansion area 934 enables users to modify the behavior of the scanner 408. Also, if there is no action modifier, then the next text element processing 900 does not perform any operation associated with action modifiers. Following the action modifier operations, a current character index is updated 936. The current character index is a pointer into the source string used to scan through the source string when performing the next text element processing 900. The next text element processing 900 then repeats operations beginning at block 904. The processing loops through blocks 904-936 until the decision 918 determines that the action is "END". When the action is "END", decision 918 causes a decision block 938 to be performed. The decision block 938 determines whether the reorder flag is set. If the reorder flag is set (see block 928), then characters within the text element are reordered 940. The reordering is preferably performed using the priority attribute which provides weighting values for different character classes. Following block 938 in the case where the reorder flag is not set or following block 940 in the case when the reorder flag is set, the next text element processing 900 is complete and returns to the Unicode converter processing 700.

FIG. 10 is a block diagram of the scanner 408. The scanner 408 among other things, includes an attributes handler 1000 and a text element handler 1002. The text element handler 1002 performs the next text element processing 900 described above with reference to FIGS. 9A and 9B. The attributes handler 1000 interacts with an attributes table 1004 to obtain the attributes for a Unicode character, as needed by the next text element processing 900 (see block 914 of FIG. 9A). The attributes include the following: direction, class, priority, symmetric swapping, subset and context. The direction attribute is used in resolving 906 direction. The class attribute is used by the scanner 408 to determine actions (e.g., ADVANCE, END). The priority attribute is used for reordering the characters within a text element (see block 940, FIG. 9B). The symmetric swapping attribute is used to whether or not symmetric swapping is needed. The context attribute is used in resolving 910 context.

FIG. 11 is a schematic diagram of a preferred format for the attributes table 1004 of FIG. 10. The attributes table 1004 includes a header portion 1100, a range table portion 1102, and an attributes table portion 1104. The header portion 1100 includes information concerning the following: total table size, checksum value, version, offset and number of elements for each table. The elements within the range table portion 1102 provide ranges in which the attributes are commonly grouped, and then for each such group provides a pointer to the appropriate portion of the attributes table portion 1104. The format of the range table portion 1102 of the attributes table 1004 includes, for each entry, a beginning of range value, an ending of range value, and a data word associated with the range. The arrangement of the attributes table 1004 facilitates compact storage of the attribute information. Alternative storage arrangements are possible but may be less efficient in terms of compactness of data storage.

FIG. 12 is a flow chart of attributes lookup processing 1200. The attributes look-up processing 1200 is performed by the attributes handler 1000, and initiated by the next text element processing 900 in block 914 of FIG. 9A.

The attributes lookup processing 1200 begins with a binary search 1202 using the ranges within the range table portion 1102 of the attributes table 1004. Once the appropriate range is identified by the binary search, the data word associated with that range is obtained 1204 from the range table portion 1102. Preferably, the first bit of the data word is an indirection bit. A decision 1206 is then made based on whether the indirection bit of the data word is set. In the case where the indirection bit of the data word is not set, the data word itself contains the attributes for the current character, therefore, the data word is returned 1208 as the attributes. On the other hand, when the indirection bit of the data word is set, then the attributes are obtained 1210 from the attributes table portion 1104 using the data word obtained from the range table portion 1102 as an index or offset into the attributes table portion 1104. Hence, the data word in this case is an index or offset to an array that contains the attributes for each character in the range. Following either blocks 1208 or 1210, the attribute lookup processing 1200 is complete and returns.

FIGS. 13A and 13B are schematic diagrams associated with a scanner table 1300 (410) utilized by the scanner 408 to determine the next action. The next action determination is invoked by block 916 of the next text element processing 900 (FIG. 9A). Recall, that the next action is used in determining the next text element. FIG. 13A illustrates a preferred format for the scanner table 1300 for use with the invention. The scanner table 1300 is a two-dimensional array having "current state" as one index and "class" as another index. The "class" refers to the class attribute. These indexes select a representative element 1302 within the scanner table 1300. FIG. 13B illustrates the representative element 1302 of the scanner table 1300. The representative element 1302 includes a next state portion 1304 containing a next state for the scanner 408, and the action portion 1306 indicates the action for the scanner 408. In effect, the scanner 408 together with the scanner table 1300 implement a state machine to determine how the scanner should operate.

The scanner table 1300 provides different next states and actions for different character encodings. FIG. 14 is a table 1400 which represents both a preferred layout and the information which would be stored in the scanner table 1300. The notations associated with the table are as follows:

Character Classes:

CC--Control character

OS--Other spacing

NS--Non-spacing

LD--Latin digit

FS--Fraction slash

JL (f)--Jamos leading consonant (filler)

JV (f)--Jamos vowel (filler)

JT--Jamos consonant trailer

NU--Not a valid Unicode character

ISS--Inhibit symmetric swapping

ASS--Activate symmetric swapping

Next States:

State 0--End, determines if text element should be returned based on states of double and half diacritics

State 1--Start state

State 2--Add non-spacing (diacritics)

State 3--check for numeric fraction

State 7--Korean Jamos

Actions:

Adv--›ADVANCE! advance to next character (the current may or may not be included in the current text element (TE)).

AdvMark--›ADVANCE+MARK! mark current character as last and advance to next character.

AdvMarkS--›ADVANCE+MARK+S! mark current character as last and advance to next character and set reorder flag.

AdvMarkASS--›ADVANCE+MARK+ASS! mark current character as last and advance to next character and activate symmetric swapping.

AdvMarkISS--›ADVANCE+MARK+ASS! mark current character as last and advance to next character and inhibit symmetric swapping.

End--End text element with last marked character.

NOTES

All functions check to see if reorder flag is set and reorder the non-spacing characters starting at start pointer. It is best to check the whole string because there may be more than one set of non-spacing marks which need to be reordered. Of course, after reordering the characters, the reorder flag is cleared.

Two examples of usage of the scanner table 1400 follow.

Input string "AAB"

The character class is OS for all three characters. The first character "A" is obtained. Beginning at the start state (state 1), the first action is AdvMark and the next state is state 2. This causes the first character "A" to be included within the current text element and causes the next character (second character "A") to be obtained. Then, at state 2, the action is End and the next state is state 0. Hence, the text element includes only the first text element. The same sequencing repeats for the second and third characters of this particular input string. Thus, each of the characters of the input string is assigned to a separate, but adjacent, text element.

Input string "A B"

The character class is OS for the first and last characters of the input string. The character class for the second character is NS because it is a combining mark. The first character "A" is obtained. Beginning at the start state (state 1), the first action is AdvMark and the next state is state 2. This causes the first character "A" to be included within the current text element and causes the next character (second character " ") to be obtained. Then, at state 2, the action is AdvMarkS and the next state is state 2. This causes the second character " " to be included in the current text element. The third character is then obtained. The action at state 2 this time is End and the next state is state 0. Hence, the text element includes the first and second characters of the input string. The third character will be placed in its own text elements as was the case with Example 1.

FIG. 15 is a flow chart of lookup text element processing 1500. The lookup text element processing 1500 is performed by the lookup handler 412 and is invoked by the block 708 within the Unicode converter processing 700 (FIG. 7).

The lookup text element processing 1500 begins by searching 1502 a variant list for an entry that matches actual attributes and requested variant. The searching 1502 of the variant is described in detail below with reference to FIG. 22. Next, a decision 1504 is made based on whether a match was found. If no match was found, then an error results 1506. On the other hand, if a match is found, then the corresponding bit mask is obtained 1508 from the variant list. Preferably, the variant list contains three fields: a variant identifier, a set of attributes, and a bit mask. If the variant identifier and the set of attributes within the variant list match the actual attributes and the requested variant, then the corresponding bit mask is chosen from the variant list. Following block 1508, the bit mask is combined 1510 with a tolerance bit mask to produce select flags. The select flags form the selection mask used in selecting the subtables of the mapping table 414 as was discussed above. Preferably, the combining 1510 is a bitwise OR operation. Next, a decision 1512 is made based on whether there is a table within the mapping table 414 for the length of the current text element. If not, an error results 1514. On the other hand, if there is a table available for the length of the current text element, the lookup table and its format are obtained 1516 for the current text element length. Then, a switch operation 1518 is performed on the format. The formats available in this implementation are: list, segment array, range and chain. If the format is a list format, then list format processing 1520 is performed. If the format is segment array, then segment array format processing 1522 is performed. If the format is range, then range format processing is performed 1524. If the format is chain, then chain format processing 1526 is performed. Following blocks 1520-1526, the result is returned 1528, thereby completing the lookup text element processing 1500.

FIG. 16 is a flow chart illustrating search variant list processing 1600. The search variant list processing 1600 is processing performed by block 1502 in FIG. 15. In other words, the search variant list processing 1600 is part of the lookup text element processing 1500 performed by the lookup handler 412 in conjunction with the mapping table 414.

The search variant list processing 1600 obtains 1602 a total count of elements in the variant list. Then, the current count is initialized to zero 1604. A decision 1606 is then made based on whether the current count is greater than or equal to the total count. If the current count is greater than or equal to the total count, then the search variant list processing 1600 returns 1608 with an error code indicating the variant not found. On the other hand, if the current count is not greater than or equal to the total count, then a decision 1610 is performed based on whether the entry in the variant list associated with the current count matches the actual attributes and the requested variant. If they do match, then the variant flags from the entry in the variant list are returned 1612. Otherwise, the current count is incremented 1614 and then the processing returns to block 1606 to continue to loop through the available variants in the variant list until either one matches or all of the variants have been considered.

FIGS. 17A and 17B are schematic diagrams of the variant list 1700. As shown in FIG. 17A, the variant list 1700 includes a variant region 1702, a desired attributes region 1704, and variant flags region 1706. FIG. 17B illustrates an actual attributes bit mask in accordance with a preferred implementation. The actual attributes bit mask 1708 is a 32-bit variable having a first portion 1710 (bits 0 and 1 ) indicating the symmetric swapping state, a second portion 1712 (bits 2 and 3) indicating vertical or horizontal forms, a third portion 1714 (bits 8 and 9) indicating resolved direction, and a fourth portion 1716 (bits 16-19) indicating context. Each bit within a portion represents a flag. The bits contain the value "0" if the flag is unknown or not set and contains the value "1" when set. The caller sets the second portion 1710, and the scanner 408 sets the first, third and fourth portions 1710, 1714 and 1716.

A desired attributes bit mask is formatted like the actual attributes bit mask, but sets the bits depending on which of the attributes is important to obtain the correct mapping for the particular table and variant (as determined by the design of the mapping table 414). The bits in the desired attributes bit mask are set to "1" for each attribute to be considered in making the mapping determination, but if all bits of a portion are set to "1", the attribute is ignored during mapping. For example, if bit 0 is "1" and bit 1 is "0", then symmetric swapping is on and will be considered when mapping is performed. On the other hand, if both bits 0 and 1 are "1", symmetric swapping is completely ignored. The remaining unused bits of the desired attributes bit mask are set to "1" and can be later assigned values if needed. A few examples of the desired attributes bit mask follow.

Suppose the direction is to be left-to-right and none of the other attributes matter. Then the desired attribute bit mask would be: xFFFFFDFF. Whereas, if the direction is to be right-to-left, with symmetric swapping on, the desired attribute bit mask would be: xFFFFFEFD. If the direction is to be right-to-left, with symmetric swapping off, the desired attribute bit mask would be: xFFFFFEFE. With each of the above different desired attribute bit mask, a different conversion code could he selected. For example, in mapping the Unicode character u0028 to MacArabic, yields x28 for the desired attribute bit mask xFFFFFDFF, xA8 for the desired attribute bit mask xFFFFFEFD, and xA9 for the desired attribute bit mask xFFFFFEFE.

FIG. 18 is a flow chart illustrating fallback handling processing 1800 in accordance with the invention. The fallback handling processing 1800 is the processing performed by the fallback handler 416 and invoked by block 716 of the Unicode converter processing 700 shown in FIG. 7.

The fallback handling processing 1800 looks-up 1802 the text element with fallback options. The lookup 1802 is similar to the lookup text element processing 1500 described above with respect to FIG. 15. The only significant difference is that now the fallback options are set so that additional subtables are considered via the select flags which have changed. Next, a decision 1804 is made based on whether conversion code for the text element was found. If conversion cede was found, then an error code is set 1806 to indicate what fallback was used in obtaining the conversion or mapping. Following block 1806, the fallback handling processing 1800 is complete and returns.

On the other hand, when the decision 1804 indicates that the conversion code is not found, then a switch operation 1808 is performed based on the fallback options. The fallback options are: default, caller defined, default followed by caller defined, or caller defined followed by default. If the fallback option is default, then the switch operation 1808 causes default processing 1810 to be performed. If the fallback option is caller defined, then the switch operation 1808 causes caller defined processing 1812 to be performed. If the fallback option is default followed by caller defined, then the switch operation 1808 causes the default processing to be performed 1814, followed by a decision 1816 and caller defined processing 1818. The decision 1816 operates to bypass the caller defined processing 1818 when a default processing 1814 is successful. If the fallback option is caller defined followed by default processing, then the switch operation 1808 causes caller defined processing to be performed 1820, followed by a decision 1822 and default processing 1824. The decision 1822 operates to bypass the default processing 1824 when the caller defined processing 1820 is successful. Following the processing associated with the switch operation 1808, a decision 1826 is made based on whether the fallback processing 1800 has successfully identified a mapping or conversion code. If the fallback processing 1800 has been unsuccessful, then a default fallback character sequence for the character set is obtained 1828. The default fallback character sequence is the conversion code(s) used when fallback lookup 1802 fails to identify a conversion code. Preferably, the default fallback character sequence is contained within the header of the mapping tables 414. For example, for ASCII, the default fallback character is typically "?". Then, following block 1828 or following decision block 1826 in the case when the fallback processing was successful in obtaining a mapping or conversion code, an error code to indicate fallback options have been used is set 1806 and the fallback handling processing 1800 is complete and returns.

FIG. 19 is a flow chart illustrating default processing 1900. The default processing 1900 is associated with the processing carried out by blocks 1810, 1814 and 1824 in FIG. 18.

A default processing 1900 initially sets 1902 a current count to zero. Next, a decision 1904 is made based on whether the current count is greater than or equal to the text element length. If it is, then the default processing 1900 is complete and returns. Otherwise, lookup processing is performed 1906 for a single Unicode character with fallback flags set. Here, the lookup is for individual characters of the text element, whereas before (block 1802) the lookup was for the entire text element. Then, a decision 1908 is made based on whether conversion code for the single Unicode character has been found. If not, then the default processing 1900 returns 1910 with an error code indicating that no individual mapping was available for the Unicode character. On the other hand, if a code conversion was found, then the current count is incremented 1912 and processing returns to block 1904 for processing of the next Unicode character within the text element.

Although FIGS. 4-19 discussed above pertain to converting from Unicode to a target encoding (From-Unicode), as mentioned above the Unicode code conversion system 300 is equally able to convert to Unicode from a different source encoding (To-Unicode). The To-Unicode processing is similar to the processing for From-Unicode but is substantially less complex. The To-Unicode processing usually does not need to scan for text elements or lookup multiple character sequences in determining the target encodings. Instead, the To-Unicode processing need only break the source string into individual characters and then find the corresponding code point in Unicode. However, in the rare case in which a mapping for a character can be influenced by the character that follows it (e.g., Indic scripts such as Devanagari, for example), scanning may be performed as discussed above.

FIG. 20 illustrates a block diagram of an embodiment of a To-Unicode code conversion system 2000 according to the invention. The Unicode code conversion system 2000 converts to Unicode (i.e., To-Unicode processing). The Unicode code conversion system 2000 includes a To-Unicode converter 2002 which receives a source string 2004 and produces a Unicode string 2006.

The To-Unicode converter 2002 performs the code conversion process via To-Unicode converter 2002 which interacts with a scanner 2008. The scanner 2008 in conjunction with the scanner table 2010 scans the source string 2004 to chunk the source string 2004 into characters. Here, unlike the From-Unicode situation, the source string is simply divided into individual characters. The To-Unicode converter 2002 then uses a lookup handler 2012 to look up the individual characters so as to obtain Unicode encodings therefor. The lookup handler 2012 uses the mapping table 2014 to obtain the character in Unicode. Additionally, the To-Unicode converter 2002 may also use a fallback handler 2016. The fallback handler 2016 operates together with the mapping table 2014 to identify one or more characters in the target encoding that are able to be used as a fallback mapping for the text element in cases where the lookup handler 2012 has been unable to identify a Unicode character.

The scanner 2008, the scanner table 2010, the lookup handler 2012, the mapping table 2014, the fallback handler 2016 and the state administrator 2018 are similar, but substantially less complex, than the corresponding devices in FIG. 4. Hence, these devices can be designed to implement both To-Unicode and From-Unicode processing. Further, when a given computer system or other electronic device is capable of performing both the To-Unicode and the From-Unicode processing, the computer system or other electronic device can operate as a hub. The hub could then operate to convert between various national character sets supported by Unicode. For example, a source national character set would first be converted to Unicode, and then converted to a target national character set.

Additional details on the code conversion system are contained in commonly-assigned U.S. patent application Ser. No. 08/527,438, entitled "UNICODE CONVERTER", which is hereby incorporated by reference.

The many features and advantages of the present invention are apparent from the written description, and thus, it is intended by the appended claims to cover all such features and advantages of the invention. Further, since numerous modifications and changes will readily occur to those skilled in the art, it is not desired to limit the invention to the exact construction and operation as illustrated and described. Hence, all suitable modifications and equivalents may be resorted to as falling within the scope of the invention.