Fontquant looks into a font and quantifies what’s in it, creating a machine-readable representation of font features that it has proven to work. It makes heavy use of the Harfbuzz shaping engine to prove the functionality of font features, rather than just looking up the feature list in the font.

The purpose of Fontquant is to:

• provide a high-level quantifiable overview of features and technical quality in order to make fonts comparable
• to make their features searchable through a user interface as part of a font library
• and for font quality assurance (QA).

If you will, the results returned by Fontquant are a kind of fingerprint of what a font can do and how it looks.

Examples:

• smcp and c2sc features are checked by applying the features to all encoded lowercase or uppercase characters in the font, using Python’s own unicodetata library to check for a character’s category to be Ll or Lu (lowercase letter or uppercase letter), and seeing whether the shaping changes after applying smcp or c2sc. The resulting value is a percentage expressed as a floating point number (0—1) that represents the total number of qualifying characters that got shaped successfully. A professional font should reach a value of 1.0 here (100%).
• Likewise, superior numbers (¹) are quantified individually. A value of 0.0 means no superior numbers are activatable through the sups feature. 0.4 means that the font contains only four superior numbers as required by legacy character sets (often ¹²³⁴), but the feature should not be advertized as fully supported until the value reaches 1.0 because a partial support is unusable in practice.

Complete documentation is available in Lib/fontquant/README.md

This described in detail the individual checks/values, along with recommendations of how to interpret them for user interfaces.

The project in Alpha stage, which means that features and implementations may change without notice.

Once we've reached Beta, we will keep a log of API changes.

Install tool with pip: pip install . on your local repository clone, or with pip install git+https://github.com/googlefonts/fontquant directly off of Github.

fontquant font.ttf. For command line options run fontquant -h.

Currently prints formatted JSON to the screen:

{
  "casing": {
    "smcp": {
      "value": 0.0
    },
    "c2sc": {
      "value": 0.0
    },
    "case": {
      "value": 0.24444444444444444
    }
  },
  "numerals": {
    "proportional_oldstyle": {
      "value": false
    },
    "tabular_oldstyle": {
      "value": false
    },
    "proportional_lining": {
      "value": true
    },
    "tabular_lining": {
      "value": true
    },
    "default_numerals": {
      "value": "proportional_lining"
    },
    "superiors": {
      "value": 0.4
    },
    "inferiors": {
      "value": 0.0
    },
    "encoded_fractions": {
      "value": 0.15
    },
    "arbitrary_fractions": {
      "value": false
    },
    "slashed_zero": {
      "value": 0.0
    }
  }
}

from fontquant import quantify

# Get results as dictionary
results = quantify("font.ttf")
# or include/run only one category:
results = quantify("font.ttf", includes=["numerals"])
# or run specific check only:
results = quantify("font.ttf", includes=["numerals/default_numerals"])
# or exclude one category, running all but that one:
results = quantify("font.ttf", excludes=["appearance"])

# Access individual check’s value
default_numerals = results["numerals"]["default_numerals"]["value"]
print(default_numerals)
>>> proportional_lining

# For variable-aware metrics (see documentation), you may define instance locations
# either manually (see below), or as "stat" (all combinations of `STAT` table definitions,
# may be a looong list), or "fvar", or "all" (both `STAT` and `fvar` combined):
results = quantify("font.ttf", includes=["appearance"], locations="wght=400,wdth=100;wght=500,wdth=100;")
# In this case, the results are indexed by their location:
print(results["appearance"]["weight"]["value"]["wdth=100.0,wght=400.0"])
>>> 0.5

• Add optional debug messages to each check to aid font QA