Scripts for generating posters that contain glyphs of every printable character in Unicode 5.1.0
C Ruby
Switch branches/tags
Nothing to show
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


These scripts require the following binaries:

   jruby    (package jruby)
   ruby     (package ruby1.8)
   curl     (package curl)
   pdftoppm (package poppler-utils)
   make     (package make)
   ocrad    (package ocrad)
   convert  (package imagemagick)

... and many from the netpbm package.  You will also need to have the
following libraries installed:


... and the font DejaVu installed as well:


You can install all of these dependencies with:

  apt-get install jruby ruby1.8 ruby curl poppler-utils make ocrad imagemagick libpng12-dev libgd-ruby1.8 ttf-dejavu-core netpbm

A completed build tree for this poster is about 18 GiB for me so
you need lots of disk space.

Step 1: Extracting the Glyphs

You should be able to run:


After a long time (at least a day on my computer) this will
generate some useful output in the directory
'individual-characters'. These files include:


    The glyph images with Unicode codepoint


    Just the glyph image, cropped from the image above


    Just the image of the Unicode codepoint number, also cropped
    from the full glyph image.

In the base directory there will also be:


    Information about the size of the -top.png image files and
    the block that the characters are from.


    This file gives the name of each image glyph and the Unicode
    codepoint acquired from doing OCR on the codepoint in the
    image.  These values will have a number of errors in them,
    since the OCR isn't perfect; if you want to be able to map
    each image to a codepoint, see the option Step 5 below.

Step 2: Drawing the Poster Strips

The script "compose-inline-breaks" will create poster strips
from these blocks.  You can invoke it like:

  ./compose-inline-breaks "Unicode 5.1.0 (Single Page Test)" 1

... where the first argument is the title of the page and the
second argument is the number of posters to spread across.  This
will generate a series of strips that we can compose into the
final posters, called something like: 


Note that these images are up to 90000 pixels wide, which
creates a number of problems.  You can view them in gimp quite
happily, but many other programs won't cope.  There are even
bugs in some graphical file managers that mean that your file
manager will crash if you open this directory.

Step 3: Composing the Poster Strips

Now, to compose the strips into a single file, just run:


That script converts each strip to PGM format and concatenates
them into gigantic PNM files, one per page of poster.

****** WARNING *****

   You need a great deal of disk space for this to run to
   completion.  The concatenated version of the poster is about
   12GB.  So, before you consider running this file you probably
   want to make sure that you have 30GB of disk free to be on
   the safe side.


The eventual output should be called (for a single page poster):


... or multiple files for multi-page posters.

In some sense this is the final output (a complete poster!) but
it's 12GB and 90000x130000 pixels, so not very useful for
printing.  To generate output you have a chance of printing see
the next section:

Step 4: Generating Printable Output

  [This is by far the worst script, but you'll probably want to
  customize it quite a bit anyway, depending on what your
  printing situation is - see Step 5.]

First, as a test, you can try generating some 600dpi A4 versions:

  ./generate-output-multi-page A4 600

If the PNG and EPS files generated from that look sensible then
you can try:

  ./generate-output-multi-page A0 600

... to generate A0 versions.

Step 5: Printing Out Your Poster

 1. Find someone with an A0 plotter.

 2. Negotiate with them at great length about whether what you
    want to do is sensible (it is :)) and what file formats they
    can deal with.

Step 1 is relatively easy, and step 2 is often very difficult.
Some printers will want to load the bitmap into Photoshop and
print it out from there.  That will probably work OK, but you
have to be very careful that they don't enable any option in the
"Print" dialog that will cause JPEG compression to be applied to
your image or the output will look horrible.  Even at A0 size
you really need the fine detail for this poster.

Many of the large HP plotters will accept TIFF files if you can
send files directly to them - this sometimes takes some
negotiation.  Before doing this, make sure that the resolution
set in the TIFF file is correct for your poster.

One other option here is to use pamdice to split the A0 image
into many A4 sheets that you can stick together.  It's much
nicer to have a proper A0 version, though.

Good luck with this step - I've found it a bit of a nightmare :/

[Optional] Step 5: Mapping Codepoints to Glyph Images

For a number of applications you might want to be able to find the
glyph image for a particular codepoint.  In order to do this, you'll
need to correct the generated codepoints.yaml file.  We can detect
most of the errors (all if we're lucky) by looking for the codepoint
values that appear to be out of place.  The script in this directory
called "test-codepoints" is designed for this.  You should do the

 # ./test-codepoints 
 Problem with: U0C00-002-00000099.png => 0x0CS3
    Not a Fixnum
 Problem with: U0E00-002-00000016.png => 0x
    Not a Fixnum
 Problem with: U1D400-002-00000165.png => 0x_D4A5
    Not a Fixnum
 Problem with: U1D400-002-00000181.png => 0x_D4B5
    Not a Fixnum
 Problem with: U1D400-002-00000191.png => 46271
    Out of order...
 Problem with: U1D400-002-00000198.png => 0x1_4C6
    Not a Fixnum
 [.. etc ..]

For each of these, do something like:

  # Look for what the codepoint should be from the image:
  xli U0C00-002-00000099.png
  # Correct it in the file:
  emacs individual-characters/non-blank/codepoints.yaml

Run ./test-codepoints again and correct any further errors,
repeating if necessary.  Once that's done you can use the script
./copy-characters-to-webspace to scale them down by 50% and copy
the glyph images to a directory where they'll have sensible

  mkdir ~/renamed-unicode-glyphs/

  ./copy-characters-to-webspace codepoints.yaml ~/renamed-unicode-glyphs/

Those files will be called things like:


... for U+00260E (BLACK TELEPHONE)


Mark Longair <>