Jacquard (wsp-jacquard)

A JSON-based configuration handler for models

Historically, a Jacquard machine is a programmable loom controlled by a chain of cards; the term "jacquard" refers to the card (or set of cards) used to configure the machine. The jacquard library is designed to facilitate application of models, where the a model's configuration (locations of data, options, parameters) are stored in a human-readable JSON file.

Jacquard is developed and maintained by WSP Canada's Systems Analytics for Policy group.

Installation

Jacquard can be installed with conda by running:

conda install -c wsp_sap wsp-jacquard

Design

One of the major design principles of the Jacquard is that model specification errors occur often, and ought to be produced in a format which is as readable as possible. Instead of getting the standard NoneType has no attribute 'iterations' or KeyError: scenario, Jacquard gives graceful messages like Item 'iterations' is missing from jacquard <model.traffic_assignment>. As a result, code which calls a Jacquard becomes self-validating. Jacquards are always ordered, and allow comments in C-style starting with // (these are stripped out during parsing).

Jacquard replaces:

from json import load

fp = r"path/to/example.json"

with open(fp) as reader:
    d = load(reader)

assert "traffic_assignment" in d, "File '%s' is missing a traffic assignment section" % fp
assert "iterations" in d['traffic_assignment'], "File '%s'.traffic_assignment is missing 'iterations" % fp
n_iterations = int(d['traffic_assignment']['iterations'])

with

from jacquard import Jacquard

config = Jacquard.from_file(r"path/to/example.json")

n_iterations = config.traffic_assignment.iterations.as_int()

Usage

The primary class in the package is the Jacquard object. Jacquards can be obtained through several class methods:

• Jacquard.from_file(fp: Path-like) reads from a JSON file. Raises JacquardParseError if it encounters an error while parsing.
• Jacquard.from_string(s: str, **kwargs) creates from an in-memory string (via json.loads).
• Jacquard.from_dict(dict_, **kwargs) creates from an in-memory dict. Dict keys are auto-converted to strings.

It is strongly recommended that JSON key names follow Python variable naming conventions and do not duplicate reserved keywords. All keys that follow the rules for Python identifiers become attributes of the returned Jacquard object.

Using a Jacquard to self-validate a file is best shown by example:

// example.json
{
  "traffic_assignment": {
    "iterations": 100,  // Set to 0 to do a shortest-path assignment
    "best_relative_gap": 0.001,
    "demand_matrix": "Projects\\TTS Demand\\auto_demand.mdf",
    "consider_background_traffic": true
  },
  "world": {
    "scenario_name": "2016 Base",
    "scenario_number": 501,
    "transit_modes": [
      "b",
      "r",
      "w"
    ],
    "random_seed": null
  }
}

from jacquard import Jacquard

config = Jacquard.from_file(r"path/to/example.json")

print('Name is:', config.name)
print('Parent is: ', config.parent)
#>> Name is: example
#>> Parent is: <root>

Each attribute returns either a child Jacquard object (one which has sub-attributes) or a JacquardValue object at the end of the tree.

...

assignment_config = config.traffic_assignment
print("Assignment sub-type", type(assignment_config))
print("Assignment namespace", assignment_config.namespace)

seed_config = config.world.random_seed
print("Random seed sub-type", type(seed_config))
print("Random seed namespace", seed_config.namespace)

# >> Assignment sub-type type<Jacquard>
# >> Assignment namespace example.traffic_assignment
# >> Random seed sub-type type<JacquardValue>
# >> Random seed namespace example.world.random_seed

JacquardValue objects expose several primitive conversion methods which handle proper type-checking:

...

n_iterations = assignment_config.iterations.as_int()
br_gap = assignment_config.best_relative_gap.as_float()
demand_matrix = assignment_jsd.demand_matrix.as_path()
bg_traffic_flag = assignment_config.consider_background_traffic.as_bool()