JIS

Derived from the documented rule rather than from the broken example: E X C E L map to U+FF25 U+FF38 U+FF23 U+FF25 U+FF2C (FULLWIDTH LATIN CAPITAL E, X, C, E, L). That is exactly the mapping batch A's ASC cases assert in the opposite direction, and exactly the value batch C asserted for DBCS("EXCEL") on the other Windows-named half of this pair. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

The Text argument description ends: "If text does not contain any half-width English letters or katakana, text is not changed." Full-width input contains no half-width letters, so the documented answer is the argument itself -- which also makes JIS idempotent, since this string is the output of the previous case. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

The page names two classes of convertible characters: "half-width (single-byte) English letters or katakana". This case covers the second. U+FF76 HALFWIDTH KATAKANA LETTER KA and U+FF85 HALFWIDTH KATAKANA LETTER NA map to U+30AB KATAKANA LETTER KA and U+30CA KATAKANA LETTER NA. Latin-only tests would never exercise this branch, and it is the branch the function is actually named after. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

"Half-width" and "full-width" describe how many BYTES a character occupied in a legacy double-byte encoding, not how many characters there are: the conversion replaces each character with one other character. So LEN is invariant under JIS even though LENB is not -- on all four builds LENB of the full-width result is 10 against LENB 5 for the argument. Asserting LEN here separates "converted the characters" from "inserted padding". Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

ASC is documented as the inverse operation ("changes full-width (double-byte) English letters or katakana within a character string to half-width (single-byte) characters"), and batch A already executed ASC's own cases. Composing the two must return the original string. This is the strongest single check in the file: it depends on nothing but the two documented rules being inverses, and it would fail if JIS converted the wrong character class, converted only some characters, or changed the length. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

Derived from the documented rule rather than from the broken example: E X C E L map to U+FF25 U+FF38 U+FF23 U+FF25 U+FF2C (FULLWIDTH LATIN CAPITAL E, X, C, E, L). That is exactly the mapping batch A's ASC cases assert in the opposite direction, and exactly the value batch C asserted for DBCS("EXCEL") on the other Windows-named half of this pair. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

The Text argument description ends: "If text does not contain any half-width English letters or katakana, text is not changed." Full-width input contains no half-width letters, so the documented answer is the argument itself -- which also makes JIS idempotent, since this string is the output of the previous case. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

The page names two classes of convertible characters: "half-width (single-byte) English letters or katakana". This case covers the second. U+FF76 HALFWIDTH KATAKANA LETTER KA and U+FF85 HALFWIDTH KATAKANA LETTER NA map to U+30AB KATAKANA LETTER KA and U+30CA KATAKANA LETTER NA. Latin-only tests would never exercise this branch, and it is the branch the function is actually named after. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

"Half-width" and "full-width" describe how many BYTES a character occupied in a legacy double-byte encoding, not how many characters there are: the conversion replaces each character with one other character. So LEN is invariant under JIS even though LENB is not -- on all four builds LENB of the full-width result is 10 against LENB 5 for the argument. Asserting LEN here separates "converted the characters" from "inserted padding". Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

ASC is documented as the inverse operation ("changes full-width (double-byte) English letters or katakana within a character string to half-width (single-byte) characters"), and batch A already executed ASC's own cases. Composing the two must return the original string. This is the strongest single check in the file: it depends on nothing but the two documented rules being inverses, and it would fail if JIS converted the wrong character class, converted only some characters, or changed the length. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

Derived from the documented rule rather than from the broken example: E X C E L map to U+FF25 U+FF38 U+FF23 U+FF25 U+FF2C (FULLWIDTH LATIN CAPITAL E, X, C, E, L). That is exactly the mapping batch A's ASC cases assert in the opposite direction, and exactly the value batch C asserted for DBCS("EXCEL") on the other Windows-named half of this pair. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

The Text argument description ends: "If text does not contain any half-width English letters or katakana, text is not changed." Full-width input contains no half-width letters, so the documented answer is the argument itself -- which also makes JIS idempotent, since this string is the output of the previous case. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

The page names two classes of convertible characters: "half-width (single-byte) English letters or katakana". This case covers the second. U+FF76 HALFWIDTH KATAKANA LETTER KA and U+FF85 HALFWIDTH KATAKANA LETTER NA map to U+30AB KATAKANA LETTER KA and U+30CA KATAKANA LETTER NA. Latin-only tests would never exercise this branch, and it is the branch the function is actually named after. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

"Half-width" and "full-width" describe how many BYTES a character occupied in a legacy double-byte encoding, not how many characters there are: the conversion replaces each character with one other character. So LEN is invariant under JIS even though LENB is not -- on all four builds LENB of the full-width result is 10 against LENB 5 for the argument. Asserting LEN here separates "converted the characters" from "inserted padding". Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.

ASC is documented as the inverse operation ("changes full-width (double-byte) English letters or katakana within a character string to half-width (single-byte) characters"), and batch A already executed ASC's own cases. Composing the two must return the original string. This is the strongest single check in the file: it depends on nothing but the two documented rules being inverses, and it would fail if JIS converted the wrong character class, converted only some characters, or changed the length. Microsoft's JIS page states the rule twice: "converts half-width (single-byte) letters within a character string to full-width (double-byte) characters", and "For Japanese, this function changes half-width (single-byte) English letters or katakana within a character string to full-width (double-byte) characters". It also warns that "The name of the function (and the characters that it converts) depends upon your language settings"; this harness runs under LC_ALL=en_US.UTF-8, a non-DBCS default language, which is stated here rather than assumed away. THE PAGE'S PUBLISHED EXAMPLES ARE UNUSABLE AS WRITTEN, so no figure is copied from them. The first reads, in the raw page source, '=JIS("EXCEL") equals "EXCEL"' -- both sides identical ASCII, i.e. the argument coming back unchanged, which directly contradicts the page's own rule since "EXCEL" is five half-width English letters. (The full-width result appears to have been flattened to ASCII somewhere in publishing; the identical defect sits on the DBCS page, which batch C recorded, and the ASC page carries the same line as its CORRECT behaviour.) The second example is rendered entirely as inline GIF images (media/fe337.gif and siblings), so no literal argument or result can be read out of it at all. Every value in this file is therefore derived from the stated rule and from the Unicode halfwidth/fullwidth block, not copied from the page.