OBD:Text encoding: Difference between revisions

From OniGalore
Jump to navigation Jump to search
(→‎Chinese: filling in missing/incorrect stuff)
m (→‎Japanese: disambig)
Line 863: Line 863:


It must be noted that, as compared to the separate .fnt files, the Japanese TSFFTahoma provides a very rudimentary implementation of JIS X 0208 (only coding for 154 double-byte glyphs, whereas the .fnt files implement 1,357) and is essentially useless/unusable.
It must be noted that, as compared to the separate .fnt files, the Japanese TSFFTahoma provides a very rudimentary implementation of JIS X 0208 (only coding for 154 double-byte glyphs, whereas the .fnt files implement 1,357) and is essentially useless/unusable.
*The Japanese engine requires all four .fnt files to be present (bails out if any of them are missing) and uses them for all of the vanilla text strings, which only contain double-byte control codes. Thus, under normal conditions, TSFFTahoma remains completely unused in the Japanese version.
*The Japanese engine requires all four .fnt files to be present (bails out if any of them are missing) and uses them for all of the vanilla text strings, which only contain double-byte control codes. Thus, under normal conditions, TSFFTahoma remains completely unused in the Japanese version, and would only be used for (artificially added) US-ASCII input.
*If the US engine is used on the Japanese game data, then the .fnt files are ignored (obviously), and the incomplete TSFFTahoma is used to render the Japanese text strings as well as the few English strings supplied by the EXE. Due to the limited character set, many strings end up broken.  
*If the US engine is used on the Japanese game data, then the .fnt files are ignored (obviously), and the incomplete TSFFTahoma is used to render the Japanese text strings as well as the few English strings supplied by the EXE. Due to the limited character set, many strings end up broken.  


Line 869: Line 869:


At the time of writing, the code points and pixel data in the Japanese .fnt files have not been thoroughly analyzed and compared with JIS X 0208. We know that 1,357 glyphs are implemented, across 27 "lead bytes" (roughly 50 ''kuten'' rows). This is much smaller than the full ''kuten'' plane, and makes sense in terms of space efficiency. We also know that some code points are non-standard (rearranged) as compared to regular Shift JIS, although we do not yet know if this rearrangement is consistent with any common variation of Shift JIS. As long as Japanese game data contains text strings that match the game's encoding, non-standard code points are not a problem (but should be kept in mind).
At the time of writing, the code points and pixel data in the Japanese .fnt files have not been thoroughly analyzed and compared with JIS X 0208. We know that 1,357 glyphs are implemented, across 27 "lead bytes" (roughly 50 ''kuten'' rows). This is much smaller than the full ''kuten'' plane, and makes sense in terms of space efficiency. We also know that some code points are non-standard (rearranged) as compared to regular Shift JIS, although we do not yet know if this rearrangement is consistent with any common variation of Shift JIS. As long as Japanese game data contains text strings that match the game's encoding, non-standard code points are not a problem (but should be kept in mind).


==Text anomalies==
==Text anomalies==

Revision as of 00:54, 5 January 2022

Originally created in English, Oni has been translated into the following seven languages: French, Italian, Spanish, German, Russian, Japanese and Chinese.

(An overview of the known language versions can be found HERE, whereas localized content is detailed HERE.)

Depending on the language version, vanilla Oni uses one of the following five encodings to render text:

  • The original US version uses a trimmed-down Mac OS Roman code page that is effectively limited to US-ASCII (96 code points).
  • European localizations (UK English, French, Italian, Spanish, German) use a custom version of Mac OS Roman (192 code points).
  • The Russian localization uses a full implementation of the Windows-1251 (Cyrillic) code page (224 code points).
  • The Chinese localization uses the EUC-CN implementation of GB 2312 (8,836 code points).
  • The Japanese localization uses 1,357 code points mostly conforming to the Shift JIS implementation of JIS X 0208.

Properties of the fonts that are eventually used to render the text (via the encoding) are briefly described throughout the page.

(A more thorough overview of the glyphs can be found HERE.)


Encodings

US English

Below is the code page implemented by TSFFTahoma in the US English version of Oni. It is based on Mac OS Roman ("MacRoman" for short), but with two differences:

  • Of the 223 printable glyphs provided by MacRoman, 42 are not implemented in TSFFTahoma (shown as grey-on-black).
  • Control point 0x7F (a typically non-printable "delete" character) has a visible box-like glyph (◻) in this implementation.
  ...0 ...1 ...2 ...3 ...4 ...5 ...6 ...7 ...8 ...9 ...A ...B ...C ...D ...E ...F
0x2... SP ! " # $ % & ' ( ) * + , - . /
0x3... 0 1 2 3 4 5 6 7 8 9 : ; < = > ?
0x4... @ A B C D E F G H I J K L M N O
0x5... P Q R S T U V W X Y Z [ \ ] ^ _
0x6... ` a b c d e f g h i j k l m n o
0x7... p q r s t u v w x y z { | } ~ ◻
0x8... Ä Å Ç É Ñ Ö Ü á à â ä ã å ç é è
0x9... ê ë í ì î ï ñ ó ò ô ö õ ú ù û ü
0xA... † ° ￠ £ § • ß ® © ™ ´ ¨ ≠ Æ Ø
0xB... ∞ ± ≤ ≥ Ұ µ ∂ ∑ ∏ π ∫ ª º Ω æ ø
0xC... ¿ ¡ ¬ √ ƒ ≈ ∆ « » …
NB
SP
 À Ã Õ Œ œ
0xD... – — ‟ ” ‛ ’ ÷ ◊ ÿ Ÿ ⁄ ¤ ‹ › fi fl
0xE... ‡ · ‚ „ ‰ Â Ê Á Ë È Í Î Ï Ì Ó Ô
0xF... Platform-Mac.png Ò Ú Û Ù ı ˆ ˜ ¯ ̆ ̇ ̊ ̧ ̋ ̨ ̌
Minor notes
  • The MacRoman layout was apparently "borrowed" before 1998, when Mac OS 8.5 came out and the international currency sign a.k.a. scarab (¤), at 0xDB, was replaced with the euro symbol (€).
  • The actual font (see HERE) has some unusual typographical features, such as a single-stroke Yen/Yuan symbol (Ұ) and a vertical-stroke cent symbol (￠).
Major notes
  • Some of the removed glyphs (most importantly ß, ù and û, but also Ê, Ú and ú) occur in common European languages. This made the US TSFFTahoma unsuitable for EFIGS localizations, requiring the creation of a new version (see below).
  • The US engine actually cannot interpret any code points beyond the US-ASCII range (first 6 rows, white background), notably failing on "…" (see "Ellipsis issue" below). This is because of a provision for Asian encoding systems (EUC-CN and Shift JIS), which use two-byte sequences starting with a high-bit byte.


European

The code page used by the five Western European versions (UK English, French, German, Spanish and Italian) is slightly different from the trimmed-down Mac OS Roman.

  • It tends to the needs of European localizations by adding back the following characters:
    German ß; French Ê and û; French/Italian ù; Spanish/Italian Ú and ú (relatively rare).
N.B. The characters Æ and ÿ are not reinstated, despite their (very rare) occurrence in French script.
  • Awkwardly enough, the six characters are not restored in their original positions (grey-on-black), but take the place of math symbols.
    Four more "math" positions are inexplicably filled with three duplicate characters (œ, ¡ and ª) and a truly enigmatic ʖ̇ , which doesn't seem to occur in any known language and has no dedicated code point in Unicode.
N.B. The broken italic font variants (see HERE) do not fully implement the 10 new glyphs and use a regular question mark instead of the ʖ̇.
  ...0 ...1 ...2 ...3 ...4 ...5 ...6 ...7 ...8 ...9 ...A ...B ...C ...D ...E ...F
0x2... SP ! " # $ % & ' ( ) * + , - . /
0x3... 0 1 2 3 4 5 6 7 8 9 : ; < = > ?
0x4... @ A B C D E F G H I J K L M N O
0x5... P Q R S T U V W X Y Z [ \ ] ^ _
0x6... ` a b c d e f g h i j k l m n o
0x7... p q r s t u v w x y z { | } ~ ◻
0x8... Ä Ç É Ñ Ö Ü á à â ä ã å ç é è
0x9... ê ë í ì î ï ñ ó ò ô ö ú ù û ü
0xA... † ￠ £ § • ß ® © ™ ´ ¨ Ø
0xB... ± Ұ µ Ê Ú ù ú û ª ß œ æ ø
0xC... ¿ ¡ ¬ ¡ ƒ ʖ̇ ª « » … À Õ Œ œ
0xD... – — ‟ ” ‛ ’ ÷ Ÿ ¤ ‹ ›
0xE... ‡ ‚ „ ‰ Â Ê Á Ë È Í Î Ï Ì Ó Ô
0xF... Ò Ú Û Ù ˆ ˜ ¯

Coincidentally, with the 10 new glyphs, the European code page has exactly 96 glyphs in the US-ASCII half and 96 in the extension half (blue).

N.B. Unlike the US version, all five Western European versions (including UK English) are able to render the full extended ASCII set.


Cyrillic

In the Russian version of Oni, TSFFTahoma implements the Windows-1251 (Cyrillic) code page, with some deviations.

  • The character 0x98, normally non-printable, is implemented as a visible box glyph (☐), slightly larger than 0x7F.
  • The character 0x81, normally a "Ѓ" glyph, is replaced with a thin space of inconsistent size (2px wide for all fonts, 3px for 13pt regular and 16pt regular).
  • The character 0xA0, normally a non-breaking space, is a space of not-so-consistent size (anywhere from single to triple width, depending on the font).
  • The character 0xAD, normally a soft hyphen, is a visible hyphen (similar to the hyphen-minus, 0x2D) for 7pt fonts, and an inconsistently sized space for other fonts.
    (Oni's engine could in theory reserve a special treatment for soft hyphens and non-breaking spaces, specified in TSFLRoman, but in practice there is no such functionality.)
  ...0 ...1 ...2 ...3 ...4 ...5 ...6 ...7 ...8 ...9 ...A ...B ...C ...D ...E ...F
0x2... SP ! " # $ % & ' ( ) * + , - . /
0x3... 0 1 2 3 4 5 6 7 8 9 : ; < = > ?
0x4... @ A B C D E F G H I J K L M N O
0x5... P Q R S T U V W X Y Z [ \ ] ^ _
0x6... ` a b c d e f g h i j k l m n o
0x7... p q r s t u v w x y z { | } ~ ◻
0x8... Ђ
S 
 P
 ‚ ѓ „ … † ‡ € ‰ Љ ‹ Њ Ќ Ћ Џ
0x9... ђ ‘ ’ “ ” • – — ☐ ™ љ › њ ќ ћ џ
0xA...
NB
SP
 Ў ў Ј ¤ Ґ ¦ § Ё © Є « ¬ – ® Ї
0xB... ° ± І і ґ µ · ё № є » ј Ѕ ѕ ї
0xC... А Б В Г Д Е Ж З И Й К Л М Н О П
0xD... Р С Т У Ф Х Ц Ч Ш Щ Ъ Ы Ь Э Ю Я
0xE... а б в г д е ж з и й к л м н о п
0xF... р с т у ф х ц ч ш щ ъ ы ь э ю я
Italic fonts
The Russian version only provides an implementation of Windows-1251 for regular and bold fonts. The five italic fonts (7pt, 9pt, 10pt, 12pt and 14pt) have exactly the same data (pixels and glyph descriptors) as for the European iteration of Mac OS Roman. This makes sense because italic fonts are inherently broken (see HERE) and thus not used by any text in vanilla Oni.
Bold 14 font
Somewhat surprisingly, the size-14 TSFT in the Russian version of TSFFTahoma does not have a complete Windows-1251 code page either. Instead it is limited to the US-ASCII character set (including the "printable delete" box at code point 0x7F), i.e., the upper section of the above table (white background). This causes no issue in vanilla Oni, but only because there is no text that uses bold 14.
Incomplete transparency
A unique "feature" of the Russian/Cyrillic TSFFTahoma is that all the characters in the extended ASCII range (0x80-0xFF) have a slightly opaque background (about 3% opacity) in the regular (non-bold) font variant. This isn't visible ingame, but only because the engine (re)posterizes all the glyphs into 4-bit grayscale when rendering (so that only opacities above 6% are visible).
Glyph alignment and spacing
Last but not least, some fonts in the Russian TSFFTahoma have inconsistent vertical alignment, the most blatant example being 12 bold: some glyphs are one pixel shorter or taller than the full line height (ascender+descender), without a properly compensated vertical glyph offset; others simply have pixels that are not properly aligned within a glyph's rectangle. Besides, many glyphs have excessive padding to the left and/or right of a character, which affects readability.
N.B. There are other examples of poor alignment, e.g. for 12 bold, the character 0x9C (њ) has its right side cut off and is thus unusable (luckily it doesn't occur in Russian script).


Chinese

The Chinese version of Oni is unique in how the main game code resides in Oni.dat, a renamed copy of the original Oni.exe from the US version that is executed indirectly by a wrapper app called oni.exe, alongside a custom text engine, xfhsm_oni.dll. The latter DLL intercepts any text about to be displayed by "Oni.dat", first reducing it to a set of two-byte control sequences, and then (if all goes well) to a set of custom glyphs, with pixel data coming from an external font file, xf_font.dat.

Unlike for the original US engine or the Japanese one, xfhsm_oni.dll does not expect any single-byte characters in the input, does not interpret US-ASCII strings in any meaningful way and never resorts to level0_Final's TSFFTahoma for text display. The pixel data comes exclusively from xf_font.dat and the expected control sequences are exclusively two-byte code points (this includes string termination; instead of a single null char xfhsm_oni.dll expects a string to end with a pair of null chars).

Unlike for other versions of Oni, the Chinese font does not have a table listing the valid code points along with their "glyph descriptors" (i.e., instructions on how to extract a glyph from the raw pixel data). Instead all the glyphs are stored as fixed-size bitmaps (16x16 pixels each) and there are exactly 94x94=8,836 glyphs, filling up a standard GB 2312 plane (qūwèi), indexed through a compact numbering scheme known as EUC-CN: each of the 94x94 code points is indexed by a pair of bytes that are both in the 0xA1-0xFE range. Code points that are not assigned under GB 2312 (e.g., rows 10-15 and 90-94) simply have blank pixel data in the corresponding regions of xf_font.dat.

Two glyph sizes are available: 16x16 glyphs are stored in the first half of xf_font.dat, and 12x12 glyphs in the second half. Each 12x12 glyph is stored in the top left corner of a 16x16 bitmap, so the row/glyph alignment is the same in both cases: 2 bytes per pixel row and 32 bytes per glyph. The pixel packing is 1-bit black-and-white (i.e., without antialiasing), much more space-efficient than the 8-bit grayscale storage used in Oni's TSFT. Another gain comes from not having any glyph descriptors (TSGAs), and from having only two fonts instead of Oni's typical 15.

All the GB 2312 glyphs listed HERE and HERE are implemented, except for the euro sign and the ten glyphs from Vertical Forms.

Unlike for other versions of Oni, an invalid code point does not interrupt the interpretation/rendering of a text string by xfhsm_oni.dll and can lead to a wide range of unexpected behavior: at best, a blank or otherwise unintended glyph will be displayed; at worst the rendered text will be garbled (memory corruption most likely), or the game may simply crash.


Japanese

Japanese Oni uses a custom two-byte encoding that is mostly consistent with Shift JIS but with some of the control sequences rearranged in seemingly non-standard ways. Like Chinese Oni, the glyph data is stored in new, external files; in this case they are .fnt files stored in GameDataFolder. Three font sizes are available, with pixel sizes 11x11 (JPN_SMALL.fnt), 12x12 (JPN_MIDDLE.fnt) and 14x14 (JPN_BIG.fnt). The 14x14 font has a bold-face variant (JPN_BOLD.fnt). All four fonts are fixed-width, i.e. all glyphs have a square bounding box.

Unlike the Chinese version, the TSFFTahoma contained in the Japanese game data is not limited to the ASCII code page. There are a total of 154 double-byte code points (Romaji, punctuation, kana and kanji) across 19 code pages (TSGA) each corresponding to a different "lead byte" (0x81, 0x82, 0x83, 0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F, 0x90, 0x91, 0x92, 0x93, 0x95, 0x96, 0x97 and 0x98).

As for the first code page of the Japanese TSFFTahoma, it implements only the 0x20-0x7F range of characters, i.e., is limited to US-ASCII. This is consistent with the simplified logic used by the Japanese engine, where any high-bit byte (in the 0x80-0xFF range) is treated as the start of a two-byte sequence. (In actual Shift JIS some high-bit bytes are interpreted as half-width kana, a feature that isn't supported by Oni's engine.)

It must be noted that, as compared to the separate .fnt files, the Japanese TSFFTahoma provides a very rudimentary implementation of JIS X 0208 (only coding for 154 double-byte glyphs, whereas the .fnt files implement 1,357) and is essentially useless/unusable.

  • The Japanese engine requires all four .fnt files to be present (bails out if any of them are missing) and uses them for all of the vanilla text strings, which only contain double-byte control codes. Thus, under normal conditions, TSFFTahoma remains completely unused in the Japanese version, and would only be used for (artificially added) US-ASCII input.
  • If the US engine is used on the Japanese game data, then the .fnt files are ignored (obviously), and the incomplete TSFFTahoma is used to render the Japanese text strings as well as the few English strings supplied by the EXE. Due to the limited character set, many strings end up broken.

It appears that the Japanese localization team initially tried to put Oni's code page system to use, and to fill in all the required JIS glyphs into TSFT and TSGA. As the number of kanji increased, supposedly, the TSFT grew prohibitively large due to the use of 8-bit grayscale storage for the pixel data, and the size taken up by the sparsely populated TSGA also increased out of proportion with the rest of the game data. At some point the engine switched to separate .fnt files, and somehow no one bothered to clean up the incomplete code pages in TSFFTahoma.

At the time of writing, the code points and pixel data in the Japanese .fnt files have not been thoroughly analyzed and compared with JIS X 0208. We know that 1,357 glyphs are implemented, across 27 "lead bytes" (roughly 50 kuten rows). This is much smaller than the full kuten plane, and makes sense in terms of space efficiency. We also know that some code points are non-standard (rearranged) as compared to regular Shift JIS, although we do not yet know if this rearrangement is consistent with any common variation of Shift JIS. As long as Japanese game data contains text strings that match the game's encoding, non-standard code points are not a problem (but should be kept in mind).

Text anomalies

Ellipsis issue

Unlike other Western versions (UK English, French, German, Italian, Spanish, Russian), the US engine treats high-bit characters as part of a two-byte control sequence (a provision for Asian encodings), and therefore fails to render any character from the extended ASCII range. This happens twice in English Oni, because the ellipsis character (…), encoded as 0xC9, was accidentally used in These Two text consoles in place of three consecutive dots (probably auto-substituted by a text editor). The result is that the two lines using a "…" are cut off at the offending character.

(A1,A0) issue

Unlike the Japanese version, where non-standard Shift JIS sequences are explicitly allowed in the .fnt files, the Chinese version does not have a code table and relies on a standard EUC-CN encoding, with exactly 8,836 code points (94x94). A proper EUC-CN control sequence consists of two bytes that are both in the range 0xA1-0xFE (single US-ASCII characters are also allowed in theory).

The text strings in the Chinese version mostly conform to the EUC-CN scheme, except for the rare occurrence of the (A1,A0) sequence. This is not a valid control sequence under any common extension of EUC-CN, and in any case it does not correspond to any pixel data within xf_font.dat, which only covers the standard 94x94 quwei plane, corresponding to a strict 0xA1-0xFE range for the two encoding bytes. Any text string including the (A1,A0) sequence is broken off at the offending character: this is known to occur for

Another illegal sequence is (0xA3,0x89), which occurs only in the SUBTmessages entry xdash1 (five identical glyphs at the end of the string).

Over-tall text

Although not strictly speaking a font issue, some of Oni's text fails to render because it doesn't fit vertically into a fixed-size frame (such as a text console). This is known to happen for These Two consoles in the English version, and possibly for other screens in other language versions.

Over-long text

Although Chinese text strings typically have a much smaller number of glyphs than English originals, this is not always the case. The Chinese glyphs are also much wider on average, with each glyph taking up 16x16 pixels, and so there are situations where the rendered Chinese line is much wider than the English original, no longer fitting on one line as intended by the context.

This is only known to cause a problem for the "resolution" item in the Options menu (a WMM_ generated at runtime). The actual dropdown list is wide enough to accommodate even the longest resolution strings, but the currently selected resolution appears in a small window that is only 150 pixels wide, too narrow even for the shortest resolution string "640×480×16位" (which needs 176 pixels). As a result the active resolution is always displayed on two lines, no longer fitting into the frame vertically and thus unreadable.

Chinese SUBT issues

The Chinese (Windows) version of Oni is unique in that no game content was actually localized except for text. Because of the relative simplicity of the task, the Chinese team did not build a new set of game data files, and merely modified the original .dat and .raw from the US version. WMDD, WMM_ and IGSt instances were patched inside each level's .dat, whereas the two SUBT files were patched in level0_Final.raw. In the case of an IGSt, text is stored in a fixed-size array (384 bytes), which has more than enough space for any translated text. WMDD and WMM_ also have fixed-size arrays (256 and 64 bytes, respectively) with at least some spare space. SUBT files, however, have a much more compact storage.

The text strings of a SUBT file (stored in level0_Final.raw and indexed from the .dat part of the SUBT) are typically packed right next to each other, separated only by a single null char. Chinese text typically uses fewer glyphs, but each glyph is taking up two bytes instead of one, including punctuation and the trailing null. Thus for short sentences or interjections it is possible for a Chinese translation to completely fill up the space used by the original string and even extend into the next entry.

None of the Chinese translations in SUBTmessages or SUBTsubtitles are actually longer than the original English text, and it is only the extra null byte that intrudes on the next entry's handle on several occasions. The affected handle essentially becomes a null string, and the corresponding subtitle is never found and displayed.

In SUBTmessages this happens only once (the message corresponding to "xf1" overwrites the first character of "xreload", so Konoko is never prompted to reload her gun in the last training room). In SUBTsubtitles there are as many as 29 anomalies, summed up in the following table.

The systematic nature of this anomaly suggests that the Chinese team were careful not to exceed the string length of the original, and merely overlooked the extra null char (and of course didn't check the ingame rendition of the subtitles all that thoroughly).

Retrieved from "https://wiki.oni2.net/w/index.php?title=OBD:Text_encoding&oldid=35337"