post-3a7ab592-4426-4caa-9eda-98517c13a3e5.py

def removeNoisyData(data):
    output = {}
    for k, v in data.items():
        if not k in SNAPSHOT and not hasattr(v, "__call__") and k != "SNAPSHOT":
            output[k] = v
    return output

def expr(s, data=None):
    if data is None:
        _data = removeNoisyData(globals())
    else:
        _data = removeNoisyData(data)

    result = s
    for name, value in _data.items():
        target = '`{}'.format(name)
        if target in result:
            result = result.replace(target, value)
    return result


def sum(expression, bottom=None, top=None, data=None):
    _expression = expr(expression, data)
    if bottom is None and top is None:
        return "\sum{{ {} }}".format(_expression)

    if bottom is None:
        return "\sum^{{ {} }} {{ {} }}".format(top, _expression)

    if top is None:
        return "\sum_{{ {} }} {{ {} }}".format(bottom, _expression)

SNAPSHOT = set(globals().keys())

def pp(s):
    print("\n{}\n".format(expr(s)))


################################################################################################################

Mq = "\mathcal{M}_{q}(t)"
Md = "\mathcal{M}_{d}(t)"
t_in_V = 't \in V'

s = lambda x: sum(x, bottom=t_in_V)

A = s(r'`Mq \cdot log \frac{`Mq}{`Md}')
B = s(r"`Mq \cdot log `Mq")
C = s(r"`Mq \cdot log `Md")

pp("`A = `B - `C")


def KL_distance(p, q, bottom=None):
    _s = lambda x, d: sum(x, bottom = bottom, data=d)
    A = _s(r'`p \cdot log \frac{`p}{`q}', locals())
    B = _s(r"`p \cdot log `p", locals())
    C = _s(r"`p \cdot log `q", locals())
    return expr("`A = `B - `C", data=locals())

Mq = "\mathcal{M}_{q}(t)"
Md = "\mathcal{M}_{d}(t)"
print(KL_distance(p = Mq, q = Md, bottom="t \in V"))