0 | ||| Measuring the visible bounds of text is incredibly hard

1 | ||| and tons of material have been written about this.

2 | |||

3 | ||| In this module, we take a pragmatic approach that produces

4 | ||| reasonable results without having to load and parse font files.

5 | ||| The drawback of this: The steps described below have to be repeated

6 | ||| for every new font we'd like to support.

7 | |||

8 | ||| A detailed introduction to typography and how fonts are specified

9 | ||| can be found [here](https://learn.microsoft.com/en-us/typography/opentype/spec/otff).

10 | |||

11 | ||| In general, we need to know the height and width of a piece of printed text

12 | ||| to properly align it with the rest of the drawing.

13 | |||

14 | ||| Text height:

15 | ||| there are different types of "height" when it comes to text, and I won't go

16 | ||| into the details here. Suffice to say that we are interested in vertically aligning

17 | ||| atom labels, charges, implicit hydrogen count and mass numbers in a

18 | ||| way that feels natural. For this, wir are mostly interested in "capHeight" of

19 | ||| a font: The height of capital letters (without descenders). This, together

20 | ||| with the Em-square size can be read from font files.

21 | |||

22 | ||| Text width:

23 | ||| While computing the height of a piece of text is non-trivial, computing its

24 | ||| width is insane. Every glyph has its own specific width, sometimes depending

25 | ||| on its neighbouring glyphs (see ligatures and kerning). Fortunately, there

26 | ||| is a quite simple method to get good approximations without being overly

27 | ||| complicated. It is described on [stackoverflow](https://stackoverflow.com/questions/16007743/roughly-approximate-the-width-of-a-string-of-text-in-python)

28 | ||| We use Python (because it has support for almost everything) to parse

29 | ||| the true type font file we are interested in and generate a dictionary

30 | ||| of the glyphs and their widths we are interested in. Using this to compute

31 | ||| the width of a piece of text at a given font size is efficient and simple

32 | ||| but not perfectly exact because it ignores kerning. It also requires large

33 | ||| dictionaries if we want to support lots of unicode characters.

34 | |||

35 | ||| Note: The Python script used to extract the glyph widths can be found in the

36 | ||| `resources` directory.

37 | module Text.Measure

38 |

39 | import Data.SortedMap

40 |

41 | %default total

42 |

43 | -- based on [stackoverflow](https://stackoverflow.com/questions/16007743/roughly-approximate-the-width-of-a-string-of-text-in-python)

44 | widths : List (Char,Bits32)

45 | widths = [('0',278),('1',278),('2',278),('3',278),('4',278),('5',278),('6',278),('7',278),('8',278),('9',278),('a',279),('b',278),('c',250),('d',278),('e',278),('f',140),('g',278),('h',278),('i',111),('j',124),('k',251),('l',111),('m',417),('n',278),('o',278),('p',278),('q',278),('r',167),('s',250),('t',139),('u',278),('v',250),('w',364),('x',250),('y',250),('z',250),('A',334),('B',334),('C',361),('D',361),('E',334),('F',305),('G',389),('H',361),('I',139),('J',250),('K',334),('L',278),('M',417),('N',361),('O',389),('P',334),('Q',389),('R',361),('S',334),('T',305),('U',361),('V',334),('W',472),('X',334),('Y',334),('Z',305),('!',139),('"',177),('#',278),('$',278),('%',445),('&',334),('\'',95),('(',167),(')',167),('*',195),('+',292),(',',139),('-',167),('.',139),('/',139),(':',139),(';',139),('<',292),('=',292),('>',292),('?',278),('@',508),('[',139),('\\',139),(']',139),('^',235),('_',292),('`',167),('{',167),('|',130),('}',167),('~',292),(' ',139)]

46 |

47 | widthMap : SortedMap Char Bits32

48 | widthMap = fromList widths

49 |

50 | averageWidth : Bits32

51 | averageWidth = sum widthMap `div` cast (length widths)

52 |

53 | charWidth : Bits32 -> Char -> Bits32