+++ /dev/null
-
-INTRODUCTION
-------------
-This document describes the process of using the GNU Unifont utilities
-to create a font. The steps described in "Using Graphical Tools" are
-more or less the steps that I (Paul Hardy) followed to add thousands
-of glyphs to GNU Unifont, except that I didn't have the luxury of just
-typing "make" to make a new font while adding those glyphs.
-
-I streamlined the font build process after I was done drawing the
-Unicode 5.1 glyphs.
-
-I know that plain ASCII text is *so* last millennium, especially for
-a package related to Unicode. Yet ASCII can be read with anything;
-hence this format.
-
-DISCLAIMER: Donald Knuth warned in his Metafont book that if someone
-started designing type, they would never again be able to look at
-a page of text normally and just read its content. There is a
-point of no return beyond which a serious font designer begins
-looking at how individual letters in a font on a page are drawn,
-and how they might be improved. Be warned!
-
-
--- Paul Hardy <unifoundry@unifoundry.com> July 2008
- with some revisions in 2013
-
-
-UNICODE
--------
-Unicode is an international standard to encode all the world's
-scripts with one universal scheme. Unicode is the default encoding
-for web pages and is gaining popularity in many other applications.
-To learn more about Unicode, look at code charts, and see the
-latest developments, check out
-
- http://unicode.org
-
-GNU Unifont follows the Unicode encoding scheme. Unicode
-defines the numeric value of a character, but does not define
-one particular font. There can be (and are) many fonts that
-support a subset of Unifont characters.
-
-In 1998, Roman Czyborra observed that there was still no font,
-free or commercial, with complete Unicode coverage. He envisioned
-a low-quality bitmapped font as an easy way to produce a font
-that covered much of the Unicode standard.
-
-
-GNU UNIFONT STRUCTURE
----------------------
-GNU Unifont is a dual-width pixel font. Roman Czyborra
-began this font in 1998 with a goal of having one glyph
-rendered for each visible character in the Unicode Basic
-Multilingual Plane (Plane 0, the first 65,536 characters).
-His original writing on this is at http://czyborra.com/unifont/.
-
-(Note that the term "character" is used very loosely here for
-simplicity; the Unicode Standard has a stricter definition
-of what constitutes a character.)
-
-The font is dual-width. Each character is 16 pixels tall, and
-either 8 or 16 pixels wide. The characters are stored in a
-unique .hex file format invented by Roman Czyborra as a convenient
-way of giving each character exactly a one line specification.
-Conversion between this .hex format and BDF font format is trivial.
-
-
-HEX FILE FORMAT
----------------
-By convention, files containing the GNU Unifont native font format
-have the extension ".hex". Their format is extremely simple, consisting
-of two fields separated with a colon (":") and ending with a newline.
-
-The first field is the Unicode code point, in hexadecimal. For all
-Plane 0 code points, this is a four digit hexadecimal number. Hexadecimal
-digits are (in order) 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E,
-and F. The Unicode Standard uses a hexadecimal number to assign
-each character a location. These locations are called "code points"
-and their range is 0 through 10FFFF, inclusive.
-
-The range 0 through FFFF, inclusive, is called the Basic Multilingual
-Plane (BMP), or Plane 0. This plane contains most characters used by
-all modern scripts of the world.
-
-Because the original goal of GNU Unifont was to cover the entire
-Unicode Basic Multilingual Plane, a four digit hexadecimal number
-suffices. In the future, I plan to allow the utilities to go beyond
-the BMP now that coverage of the Unicode BMP is finally complete.
-
-The programs will probably handle glyphs beyond the BMP properly, but
-that capability is considered experimental, as the focus was to cover
-the BMP. The C programs allow up to an eight digit hexadecimal code
-point, but not all the utilities have been thoroughly tested beyond
-the BMP.
-
-The second field is a string of hexadecimal digits. There are 32
-digits for a character that is 8 pixels wide, and 64 digits for a
-character that is 16 pixels wide.
-
-The good news is you don't have to worry about these long digit
-strings. Roman Czyborra wrote a utility, "hexdraw", to convert
-.hex fonts to a form that can be edited with a plain text editor,
-then converted back into .hex format.
-
-Paul Hardy wrote two utilities to do the same thing except with
-bitmapped graphics images for editing with a graphics editor:
-"hex2bmp" converts a block of 256 characters into a graphics
-file, and "bmp2hex" converts such a graphics file back into
-.hex format. These bitmaps display the 256 characters in a block
-arranged in a 16 by 16 character grid. The graphics editor must
-maintain the image as a monochrome (black and white) file, with
-one bit per pixel. After conversion from a .bmp file back to
-a .hex file, the next step is conversion to a BDF file and BDF
-can only encode a pixel being on or off (i.e., black or white
-only with no intermediate shades of gray).
-
-
-USING GRAPHICAL TOOLS
----------------------
-Let's use an example. Suppose you want to modify the Coptic letters
-in the range U+2C80..U+2CFF ("U+" is Unicode shorthand). These
-letters are in the upper half of the block U+2C00..U+2CFF. The
-Unicode utilities in this package refer to this as "page" 2C.
-("Page" is not a Unicode term -- it is just a term unique to this
-package to refer to a block of 256 code points/characters).
-
-The steps to follow will be:
-
- 1) Convert .hex version of the page 2C range as a 16 by 16
- bitmapped grid.
- 2) Modify the bitmap in any graphics editor, being careful
- to re-save it as a Windows Bitmap (.bmp) or Wireless
- Bitmap file when finished.
- 3) Convert the modified bitmap back into a .hex font file.
- 4) Merge the results with the original unifont.hex file.
- 5) Run "unidup" on the resulting file to guard against
- duplicate character definitions.
- 6) Create the new bitmapped version of the font.
- 7) Check the compiled font for duplicates.
- 8) If there are duplicates, remove them and go back to
- Step 5.
- 9) Create the new TrueType version or other versions of the font.
-
-Step 1: Convert the .hex range into a bitmap grid.
- Assuming our font file is named "unifont.hex", type
-
- unihex2bmp -p2C < unifont.hex > uni2C.bmp
-
-Step 2: Modify uni2C.bmp with your favorite graphics editor. Note
- that whatever graphics editor you use must preserve the
- file as a black and white bitmap (monochrome), with one bit
- per pixel. During editing, you can draw guidelines outside
- the actual 16x16 font pixel area; they will be ignored when
- converting back into .hex format. You can also erase the
- grid borders between code points on purpose or by accident,
- and it will have no effect on the generated .hex file. Just
- don't erase the code point numbers on the outer edges of
- the grid. The conversion from .bmp back to .hex only looks
- at the "U+...." and other code point numbers, and at each
- code point's 16x16 pixel area. Every other artifact in the
- final graphics file outside these areas is ignored.
-
- If a new version of Unicode adds glyphs to a page that were
- previously unassigned, be sure to remove the newly-assigned
- code point from the "unassigned.hex" file because the code
- point is no longer unassigned.
-
-Step 3: Convert the edited .bmp file back into .hex format:
-
- unibmp2hex < uni2C.bmp > uni2C.hex
-
- Note that the conversion from a bitmap image to a .hex file
- can't distinguish between a legitimate single- or double-width
- space character and a code point that does not have an assigned
- value. Therefore, space glyphs are separately contained in the
- spaces.hex file.
-
-Step 4: Merge the results with the original "unifont.hex" file.
- This requires several sub-steps:
-
- - Edit the original "unifont.hex" file and delete the
- lines that begin with "2C".
- - Insert the "uni2C.hex" file into "unifont.hex", either
- with a text editor such as vi or emacs, or with a
- Unix command such as:
-
- sort uni2C.hex unifont.hex > new-unifont.hex
-
- This second option (using "sort") is preferred, because
- "unidup" (in Step 5) might miss duplicate code points
- if your final result isn't in proper order.
-
-Step 5: Make sure there are no duplicates with "unidup":
-
- unidup < unifont.hex
-
- or
-
- unidup < new-unifont.hex
-
- depending on the name of your final font file. If there
- is no output, your modified font contains no duplicates.
-
- This editing is best done on an input .hex file, such as
- "unifont-base.hex".
-
-Step 6: Create the new bitmapped version of the font. In the
- font/ directory, type
-
- make hex
-
-Step 7: Check the compiled font for duplicates. Change to the
- font/compiled directory and run
-
- unidup < mynewfontfile.hex
-
- for whatever font file you created.
-
-Step 8: If there are duplicates, remove them in the font/ directory
- and go back to Step 5.
-
-Step 9: Create the new TrueType version of the font and all other
- bitmapped versions. From the font/ directory, type
-
- make distclean && make
-
- Then be prepared to wait a long time unless you are using
- a computer with plenty of RAM and CPU horsepower. Your
- computer should have at least 256 Megabytes of virtual
- memory (RAM), and at least 250 Megabytes of free disk space.
-
- To only create a BDF font, in the font/ directory just type
-
- make bdf
-
- To only create a BDF and PCF font, in the font/ directory type
-
- make pcf
-
- Creating a BDF font is the first step in creating a PCF font
- (not counting generating the compiled master ".hex" input file).
- BDF fonts can be created just with the tools in this package.
- PCF fonts are created by running "bdftopcf" on the BDF font.
- TrueType fonts require FontForge.
-
-
-USING HEXDRAW
--------------
-Roman Czyborra's original utility to edit glyphs is the "hexdraw"
-Perl script. Using the same script as above, Coptic, here are the
-steps for modifying "unifont.hex" using "hexdraw".
-
-First, realize that GNU Unifont has tens of thousands of glyphs
-(characters, using the term character loosely). In this example,
-out of the tens of thousands of glyphs, we want to modify the range
-U+2C80..U+2CFF (only 128 glyphs).
-
-The range U+2C80..U+2CFF could be extracted from unifont.hex by
-using the "egrep" utility to look for lines beginning with "2C8"
-through "2CF", or that range could be isolated by copying "unifont.hex"
-into another file, and deleting all lines except the desired range.
-
-The following steps will probably minimize typographical errors,
-but they aren't the only way.
-
- 1) "grep" the desired block of 256 glyphs and convert them
- into a text representation for editing.
- 2) Edit the block with a text editor, such as vi or emacs.
- 3) Convert the edited text file back into .hex format.
- 4) Delete the edited block range from the original font file.
- 5) Merge the two .hex files into one file.
- 6) Check for duplicates with "unidup".
- 7) Generate new fonts as described in the Using Graphical
- Tools section above.
-
-Step 1: Extract the desired block with grep:
-
- grep "^2C" unifont.hex | hexdraw > uni2C.txt
-
-Step 2: Edit "uni2C.txt" with a text editor.
-
-Step 3: Convert the text file back into .hex format:
-
- hexdraw < uni2C.txt > uni2C.hex
-
-Step 4: Delete the lines in the original "unifont.hex" file that
- begin with "2C".
-
-Step 5: Merge the two files:
-
- sort unifont.hex uni2C.hex > new-unifont.hex
-
- or use Roman's "hexmerge" utility:
-
- hexmerge unifont.hex uni2C.hex > new-unifont.hex
-
-Step 6: Check for duplicates:
-
- unidup < new-unifont.hex
-
- Of course, remove any reported duplicates.
-
-Step 7: Build the font as in the Using Graphical Tools section
- above. This can be as simple as typing
-
- make
-
- in the font/ directory.
-
-I (Paul Hardy) had only used hexdraw in the very beginning of my
-work on the GNU Unifont. Once I got my graphics programs working,
-I ignored it for months. Then I wanted to go through all of the
-Yi Syllables and Yi Radicals -- over 1000 glyphs -- to fine-tune
-their horizontal alignment after I drew them. hexdraw turned out
-to be the perfect tool for this. By printing hyphens ("-") as
-place holders where a pixel is off, it allowed me to verify space
-to the left and right of each character. I later used hexdraw
-for similar fine-tuning of spacing on Hangul and other glyphs.
-It is ideal for the task.
-
-
-CHECKING COVERAGE
------------------
-There should never be duplicates in a .hex file. If there are, remove
-them before the .hex font is turned into a .bdf or other font. The
-notes above include making liberal use of "unidup" to avoid such a
-situation.
-
-The "unipagecount" program will print a hexadecimal number of code
-points that have coverage within each 256 code point block. The
-hexadecimal number will therefore range from 0 (no coverage) to
-100 (= 256 decimal; full coverage). If a number is ever more than
-100 hexadecimal, there's an extra character (glyph) for that page.
-
-To further look at the coverage within just one 256 code point
-page (using page 2C, containing Coptic, as our example) use
-
- unipagecount -p2C < unifont.hex
-
-Note that the "page number" can use upper- or lower-case letters:
-"-p2C" or "-p2c" will both work. That will print a 16 x 16 grid
-of U+2C00..U+2CFF. Of course, without placeholder glyphs for the
-unassigned code points from "unassigned.hex" in the U+2C00..U+2CFF
-range, unipagecount can give a lower number than the true coverage.
-
-Using the "-l" flag with "unipagecount" will produce an HTML
-table with links to corresponding graphics images. You can get
-an idea of how this works in the "./font/compiled" directory after
-running "make"; the "index.html" file in that directory will have
-a table with links to the 256 glyph maps in the ./font/compiled/bmp
-subdirectory.
-
-With unipagecount, the background color of the cells will range from
-red (for 0% complete in that 256 code point block) to orange
-(a little coverage) to yellow (more coverage) to green
-(complete coverage). If a cell looks light red or pink,
-the corresponding code page might accidentally have duplicate
-characters. Verify that with "sort unifont.hex | unidup"
-(substituting the name of your .hex file for "unifont.hex").
-
-To see the coverage of each Unicode script, copy the file "./coverage.dat"
-into the same directory as the "unifont.hex" file you're examining.
-Then run
-
- unicoverage < unifont.hex > coverage.out
-
-This will give you all the scripts within the Unicode Basic
-Multilingual Plane, in order, with a percentage (0.0% through
-100.0%) of each script's coverage. Note that to get the true
-coverage of assigned code points, you'll have to merge unassigned.hex
-with the rest of unifont.hex if not done by default in your setup.
-
-Using the .hex files in font/hexsrc, you can create a font with
-all available glyphs with
-
- sort font/hexsrc/*.hex >unifont-whole.hex
-
-and run unicoverage using the resulting "unifont-whole.hex" file.
-
-
-CUSTOM BUILDS
--------------
-The font can be built from within the "font/" directory by simply typing
-
- make
-
-Typing "make" from the top-level directory (one level above the "font/"
-directory) will also build the font.
-
-By default, source glyphs are read from the "font/hexsrc/" directory.
-Glyphs for unassigned code points are built into the font by default,
-but glyphs for Private Use Area code points are not built into the font.
-
-All of the .hex file names can be replaced selectively. The list of
-component hex file variables is:
-
- UNIFONTBASE - The bulk of Unifont scripts
- CJK - Most of the CJK Ideographs
- HANGUL - Hangul Syllables block
- SPACES - Space glyphs, single- and double-width
- UNASSIGNED - Glyphs for unassigned code points
- PUA - Glyphs for the Private Use Area
-
-So, for example, to build a font that includes four-digit hexadecimal
-code point glyphs (as white on black) for the Private Use Area, type
-
- make PUA="hexsrc/pua.hex"
-
-To build a font that includes your own special PUA glyphs, type
-
- make PUA="mycoolPUA.hex"
-
-or whatever the name of your PUA glyph .hex file is named.
-
-To create a build that includes the supplied PUA glyphs but not the
-Unassigned glyphs, type
-
- make PUA="hexsrc/pua.hex" UNASSIGNED=""
-
-If you create a custom font build of your own in one gigantic file,
-you can build with just this file by declaring all the ordinary files
-to be null:
-
- make UNIFONTBASE="mycustomfont.hex" CJK="" HANGUL="" UNASSIGNED="" PUA=""
-
-Note that this command did not include an override for the glyphs for
-spaces contained in "font/hexsrc/spaces.hex"; that is, the variable
-SPACES was not redefined on the command line. You could also pass the
-argument SPACES="", but just be aware that those spaces glyphs are in
-a separate file for a reason. When graphical (".bmp") glyph files are
-converted back into hex string (".hex") format, the "unibmp2hex" utility
-doesn't know if a blank glyph area is a space glyph or not, so it doesn't
-encode anything. The file "font/hexsrc/spaces.hex" contains glyphs that
-are strings of 0s, with length depending on whether the space is nominally
-a single- or double-width space.
-
-
-SEEING THE BIG PICTURE (LITERALLY!)
------------------------------------
-The Unifont 6.3 release adds a new program, unifontpic. This
-produces a chart of all the Basic Multilingual Plane glyphs in
-Unifont. By default the chart is arranged as a 256-by-256 glyph
-table. Specifying the '-l" option produces a chart that is
-16 glyphs wide by 4,096 glyphs long. See unifontpic(1) for more
-information.
-
-To generate a chart with all your glyphs in one giant unifont.hex
-file, type the command
-
- unifontpic < unifont.hex > unifont.bmp
-
-The output is a monochrome Bitmap Graphics Format (.bmp) file.
-If you prefer PNG files, use your favorite graphics program or
-conversion program to convert the BMP file to a PNG file.
-
-This utility is especially useful if you're customizing the font,
-for example adding your own Private Use Area glyphs.
-
-The default 256-by-256 code point chart will render satisfactorily
-on a sheet of paper approximately 3 feet by 3 feet at 120 dots per
-inch. Thus the square format is suitable for printing.
-
-The "long" version, created with "unifontpic -l", only produces
-16 glyphs per row. This is more useful for scrolling through on
-a computer screen. GIMP can read the resulting .bmp file (at least
-under GNOME), but not all utilities can. The output file is over
-65,536 pixel rows tall, which causes problems with some graphics
-programs.
-
-
-COMBINING CIRCLES
------------------
-The earliest versions of Unifont (before the 5.1 release) used glyphs
-that showed dashed circles for combining characters. This is how the
-glyphs appear in The Unicode Standard code charts. In version 5.1,
-Paul Hardy was able to get combining characters to appear superimposed
-correctly in the TrueType version of the font (note: this won't work
-in the BDF or PCF versions).
-
-With combining characters working, Paul Hardy created variations of
-Unifont with and without combining circles, the idea being that the
-version without combining circles would be used to create the TrueType
-font. The variation with combining circles was kept for reference.
-
-Version 6.2 simplified the build to produce only one font variation,
-without combining circles.
-
-Version 6.3 introduces a new utility, unigencircles, to superimpose
-combining circles over glyphs with code points in a combining.txt file.
-
-To run unigencircles, start with the file "font/ttfsrc/combining.txt"
-and type a command of this form:
-
- unigencircles combining.txt < unifont.hex > unifont-circles.hex
-
-where "unifont.hex" is a single file containing all the glyphs you
-wish to render.
-
-Because the output is another .hex file, you can use all Unicode
-utilities with this file just as you would with the .hex files in
-"font/hexsrc".
-
-If you want to build a font from this resulting file, you could type
-
- make UNIFONTBASE="unifont-circles.hex" CJK="" HANGUL="" UNASSIGNED="" PUA=""
-
-as discussed above. In this case, if you included "font/hexsrc/spaces.hex"
-in the "unifont.hex" input file, you should also set SPACES="" on the
-command line so that you only read in your final custom "unifont-circles.hex"
-file.
-
-
-INSTALLING FONTS ON UNIX/LINUX
-------------------------------
-Compress PCF fonts using "gzip -9 fontname.pcf". Copy the resulting
-"fontname.pcf.gz" file to /usr/share/fonts/X11/misc/ or place in a
-local font directory if your windowing software supports that (for
-example, ~/.fonts/).
-
-Copy TrueType fonts to /usr/share/fonts/truetype uncompressed or place
-in your local font directory. Note: on some versions of Unix, such as
-Solaris, the name of the TrueType directory might be TrueType and
-might be under /usr/share/fonts/X11 -- examine the system fonts
-directories, then modify the font "make install" rule accordingly.
-
-On most flavors of Linux with the latest "xset" utility (including
-the latest Debian and Red Hat releases), the following command should
-allow you to start using the font immediately:
-
- xset fp rehash
-
-The safest way to make sure the system knows about the new fonts will
-be to restart the X Window System or even reboot.
-
-
-INSTALLING FONTS ON MICROSOFT WINDOWS
--------------------------------------
-Copy the TrueType (*.ttf) font into your C:\WINDOWS\Fonts folder.
-
-
-INSTALLING FONTS ON MACS
-------------------------
-Copy the TrueType (*.ttf) font into the Fonts folder under the
-Finder folder.
-
-
-CREATING A BRAND NEW FONT
--------------------------
-These tools will only produce glyphs that are 16 pixels tall, and
-8 or 16 pixels wide. This current release is furthermore only fully
-tested over the Unicode Basic Multilingual Plane (code points from
-U+0000 through U+FFFF, with none above that range -- this will change
-in the future). The tools can be used to create and manipulate
-any bitmapped font within those constraints.
-
-To create a brand new font (or even to add a new script to GNU Unifont
-in the future), plan out the basic dimensions of the characters:
-
- - How far above the lowest pixel will the baseline appear (in
- other words, how many rows are necessary for descenders)?
-
- - How many pixels must be empty on top and bottom for accents
- (in other words, what will the height of capital letters be)?
-
- - Must glyphs be centered, left-aligned, or right-aligned?
-
- - For a Latin font, what will the "x-height" be (the height of
- a lower-case "x" and related letters, such as "n" and "r")?
-
-Look over the entire script and plan out a grid that is consistent
-for the entire script.
-
-GNU Unifont characters for the most part leave the right-most column
-of pixels blank if possible for left-to-right scripts. Centering is
-done around the fourth pixel (if a glyph is eight pixels wide) or
-around the eighth pixel (if a glyph is 16 pixels wide).
-
-Consistent cap heights, x-heights, descender depths, and centering
-will produce a better looking font.
-
-Experimenting and (above all) having fun with these utilities is
-the best way to learn.
-
-
-UPDATES TO UNICODE
-------------------
-If a currently unused code point is assigned to a character
-in the future, the font can be updated as follows:
-
- 1) Delete the code point's entry from "unassigned.hex",
- as that code point will no longer be unassigned.
-
- 2) Determine which existing .hex file should contain the
- newly defined character (examine the files to see
- which one contains other glyphs in the script.
-
- - "unifont-base.hex" contains most scripts
- - "wqy.hex" contains most CJK ideographs
- - "hangul-syllables.hex" contains the Hangul
- Syllables block
-
- If in doubt (for example, if a new script is added to
- Unicode and you're not sure which .hex file to augment),
- add the new glyphs to "unifont-base.hex".
-
- 3) Update the appropriate .hex file.
-
- 4) Consider if "./font/coverage.dat" has to be updated.
- Follow its existing format to insert a new script,
- being sure to change any previously unassigned ranges
- before or after the new script.
-
- 5) Make a new .hex version of the font, and verify that
- you didn't introduce any duplicates.
-
- 6) Run "unipagecount" and/or "unicoverage" as described
- above to verify that you have not mistakenly deleted
- any existing characters.
-
-Enjoy!
-
git.shiar.nl / unifont.git / blobdiff