我正在尝试组合 seq-m 和 error-m 来对可能返回错误的事物进行列表推导。我的输出有意想不到的类型，但除此之外它实际上似乎是合理的。我在下面分解了我的代码，但这里有一个 working gist

这是我的 monadic 业务逻辑

def get_loan(name):
    m_qualified_amounts = (
           bind(get_banks(name), lambda bank:
           bind(get_accounts(bank, name), lambda account:
           bind(get_balance(bank, account), lambda balance:
           bind(get_qualified_amount(balance), lambda qualified_amount:
                    unit(qualified_amount))))))
    return m_qualified_amounts

names = ["Irek", "John", "Alex", "Fred"]
for name, loans in zip(names, map(get_loan, names)):
    print "%s: %s" % (name, loans)

输出

Irek: [None, 'Insufficient funds for loan, current balance is 35000', None, 'Insufficient funds for loan, current balance is 70000', None, 'Unable to get balance due to technical issue for Wells Fargo: 3']
John: [None, 'Insufficient funds for loan, current balance is 140000']
Alex: [[245000], None, [280000], None]
Fred: (None, 'No bank associated with name Fred')

我希望看到元组列表——该列表是列表推导的结果，最终列表中的每一项都应该是错误单子(monad)(value, error 元组)中的一个值。就好像 seq_bind 移除了太多层级的嵌套。

这是我对 monad 的定义，如果它不正确，它非常接近，因为两个 monad 都是独立工作的，只是没有结合。

def success(val): return val, None
def error(why): return None, why
def get_value(m_val): return m_val[0]
def get_error(m_val): return m_val[1]

# error monad
def error_unit(x): return success(x)
def error_bind(mval, mf):
    assert isinstance(mval, tuple)
    error = get_error(mval)
    if error: return mval
    else: return mf(get_value(mval))

def flatten(listOfLists):
    "Flatten one level of nesting"
    return [x for sublist in listOfLists for x in sublist]    

# sequence monad
def seq_unit(x): return [x]
def seq_bind(mval, mf):
    assert isinstance(mval, list)
    return flatten(map(mf, mval))

# combined monad !!
def unit(x): return error_unit(seq_unit(x))
def bind(m_error_val, mf):  
    return error_bind(m_error_val, lambda m_seq_val: seq_bind(m_seq_val, mf))

单一API

def get_banks(name):
    if name == "Irek": return success(["Bank of America", "Wells Fargo"])
    elif name == "John": return success(["PNC Bank"])
    elif name == "Alex": return success(["TD Bank"])
    else: return error("No bank associated with name %s" % name)

def get_accounts(bank, name):
    if   name == "Irek" and bank == "Bank of America": return success([1, 2])
    elif name == "Irek" and bank == "Wells Fargo": return success([3])
    elif name == "John" and bank == "PNC Bank": return success([4])
    elif name == "John" and bank == "Wells Fargo": return success([5, 6])
    elif name == "Alex" and bank == "TD Bank": return success([7, 8])
    else: return error("No account associated with (%s, %s)" % (bank, name))

def get_balance(bank, account):
    if bank == "Wells Fargo":
        return error("Unable to get balance due to technical issue for %s: %s" % (bank, account))
    else:
        return success([account * 35000])  #right around 200,000 depending on acct number

def get_qualified_amount(balance):
    if balance > 200000:
        return success([balance])
    else:
        return error("Insufficient funds for loan, current balance is %s" % balance)

也在寻找改进代码的方法。标记为 haskell 和 clojure，因为这在这些语言中是惯用的，python 社区对此不感兴趣。

最佳答案

在 Haskell 中使用 Monad Transformers 像这样通过堆叠组合 monad .暂时搁置 Daniel Wagner 的 ListT 不是 monad 的观点。您有两个具有类型的 monad:

1. 列出一个，它看起来像[x,y,z]
2. (Error e) a 看起来是 x, None 或 None, err

如果将一个转换为 monad 转换器并组合它们，有两种方法:

1. (ErrorT e) 列出一个，看起来像 [ (x,None), (y,None), (None, err) ]
2. ListT (ErrorT e) a 看起来像 [x,y,z], None 或 None, [x,y,z]

你想要一个对列表，所以我想你想要第一个形式。但是您的简单测试不同意这一点。您的 unit 不会返回 (1.) 中的成对列表，而是返回一对列表和 None，即 (2.)。

所以你要么有倒退的东西，要么你有一个更复杂的单子(monad)。我会尝试将您的要点修改为 (1.)。

我认为这段代码可能会做你想做的事:

def flatten(listOfLists):
    "Flatten one level of nesting"
    assert isinstance(listOfLists, list)
    if len(listOfLists) > 0:
        assert isinstance(listOfLists[0], list)
    return [x for sublist in listOfLists for x in sublist]

# sequence monad
def seq_unit(x): return [x]
def seq_bind(mval, mf): return flatten(map(mf, mval))

# Decompose ErrorT e m a
def get_value(m_val): return m_val[0]
def get_error(m_val): return m_val[1]

# hard coded "(ErrorT e) List a" instance of throwError, note that seq_unit is hardcoded
def error_throwError(err): return (None, err)
def errorT_list_throwError(err): return seq_unit(error_throwError(err))

# "(ErrorT e) List a" monad
def error_unit(x): return (x,None)
def errorT_list_unit(x): return seq_unit(error_unit(x))

def error_bind(mval, mf):
    assert isinstance(mval, tuple)
    error = get_error(mval)
    if error:
        return error_throwError(error)
    else: 
        return mf(get_value(mval))

# Cannot have multi-line lambda
def errorT_list_bind_helper(mval, mf):
    assert isinstance(mval, tuple)
    error = get_error(mval)
    if error:
        return errorT_list_throwError(error)
    else: 
        return mf(get_value(mval))

def errorT_list_bind(mval, mf): return seq_bind(mval, lambda v: errorT_list_bind_helper(v, mf))

# combined monad !! (ErrorT e) List a
unit = errorT_list_unit
bind = errorT_list_bind
throwError = errorT_list_throwError

# hard coded "lift :: List a -> (ErrorT e) List a"
def lift(mval):
    assert isinstance(mval, list)
    # return [ (val,None) for val in mval ]
    # return [ errorT_list_unit(val) for val in mval ]
    return seq_bind(mval, lambda v : unit(v))

def get_banks(name):
    if name == "Irek": return lift(["Bank of America", "Wells Fargo"])
    elif name == "John": return unit("PNC Bank")
    elif name == "Alex": return unit("TD Bank")
    else: return throwError("No bank associated with name %s" % name)

def get_accounts(bank, name):
    if   name == "Irek" and bank == "Bank of America": return lift([1, 2])
    elif name == "Irek" and bank == "Wells Fargo": return unit(3)
    elif name == "John" and bank == "PNC Bank": return unit(4)
    elif name == "John" and bank == "Wells Fargo": return lift([5, 6])
    elif name == "Alex" and bank == "TD Bank": return lift([7, 8])
    else: return throwError("No account associated with (%s, %s)" % (bank, name))

def get_balance(bank, account):
    if bank == "Wells Fargo":
        return throwError("Unable to get balance due to technical issue for %s: %s" % (bank, account))
    else:
        return unit(account * 35000)  #right around 200,000 depending on acct number

def get_qualified_amount(balance):
    if balance > 200000:
        return unit(balance)
    else:
        return throwError("Insufficient funds for loan, current balance is %s" % balance)

# monadic business logic
def get_loan(name):

    m_qualified_amounts = (
           bind(get_banks(name), lambda bank:
           bind(get_accounts(bank, name), lambda account:
           bind(get_balance(bank, account), lambda balance:
           bind(get_qualified_amount(balance), lambda qualified_amount:
                    unit(qualified_amount))))))

    assert isinstance(m_qualified_amounts, list)
    assert isinstance(m_qualified_amounts[0], tuple)
    return m_qualified_amounts

names = ["Irek", "John", "Alex", "Fred"]

for name, loans in zip(names, map(get_loan, names)):
    print "%s: %s" % (name, loans)

输出是

Irek: [(None, 'Insufficient funds for loan, current balance is 35000'), (None, 'Insufficient funds for loan, current balance is 70000'), (None, 'Unable to get balance due to technical issue for Wells Fargo: 3')]
John: [(None, 'Insufficient funds for loan, current balance is 140000')]
Alex: [(245000, None), (280000, None)]
Fred: [(None, 'No bank associated with name Fred')]

关于python - 结合 maybe 和 seq monads : confused at the output，我们在Stack Overflow上找到一个类似的问题： https://stackoverflow.com/questions/10059163/