gunthercox / mathparse Goto Github PK

As I run the chatterbot with math logic and evaluate the following sentences:

“10 plus 2 squared times 3”
“10 plus (2 squared) times 3”

must have the same results with 22.

However, the previous answer is 2,985,984, and the latter one is 22.

I think it happens because of mathparse, especially to_postfix() (the precedence of ^ and others)

Only a few of the ISO 639-2 language codes are currently supported.

• https://www.loc.gov/standards/iso639-2/php/code_list.php

• Include test to show that an equation containing sqrt or square root can be evaluated.

Example: "fifty three plus nine"

I am having issues with words like seventeen being evaluated as 7teen.

With python 3.6+, would it help to list the English numbers like so:

    'numbers': {
        'eleven': 11,
        'twelve': 12,
        'thirteen': 13,
        'fourteen': 14,
        'fifteen': 15,
        'sixteen': 16,
        'seventeen': 17,
        'eighteen': 18,
        'nineteen': 19,
        'twenty': 20,
        'thirty': 30,
        'forty': 40,
        'fifty': 50,
        'sixty': 60,
        'seventy': 70,
        'eighty': 80,
        'ninety': 90,
        'zero': 0,
        'one': 1,
        'two': 2,
        'three': 3,
        'four': 4,
        'five': 5,
        'six': 6,
        'seven': 7,
        'eight': 8,
        'nine': 9,
        'ten': 10
    },

Running mathparse.replace_word_tokens('fifty thousand + 1', language='ENG') return 5(0 * 1000). I can look into fixing this, but it appears that it was hardcoded somewhere that the number before a scale is always one digit long.

• Add test to assert "Fifty three point four times seven" == "53.4 * 7"

The following line doesn't get the right answer.
ret = mathparse.parse('one hundred times fifty four', language='ENG')
I found that function replace_word_tokens in mathparse.parse goes wrong.Take one hundred times fifty four for example,we expect get '(1 * 100) * (50 + 4)' but get '(1 * 100) * (50 4)'.There is always a '+' missing when it comes to situations like 'twenty two','thirty three' and etc.

Love this library, could be very useful. Mathparse could also make a good alternative to the pricey Wolfram Alpha python API because of the natural language input that translates into mathematical functions, if you were to add support for Integration, summations, solving, graphing, etc.

Example: "one hundred plus two hundred"

The tokenizer currently splits math tokens based on whitespace.

In Mandarin 二加二 means "'two plus two" but no spaces are used between the characters.

The tokenizer needs to be modified so that it will split these values correctly.

I believe the optimal solution might be to have the tokenizer ignore whitespace and traverse the values in the string based on the type they get recognized as. So, in this case, the first value would be read 二 and added to the stack of numbers, then the next value would be read 加 and added to the binary operator stack, and so on. No need to split on whitespace.

Previously reported as gunthercox/ChatterBot#1115

Taken from the examples in the README, if we run

mathparse.parse('50 * (85 / 100)', language='FRE')

we get the error

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "mathparse/mathparse.py", line 246, in parse
    string = replace_word_tokens(string, language)
  File "mathparse/mathparse.py", line 88, in replace_word_tokens
    operators.update(words['unary_operators'])
KeyError: 'unary_operators'

It appears we just need to add unary operators to every language or check if the language has unary operators before calling

operators.update(words['unary_operators'])

If an input such as '42 / 0' is received, this should be handled cleanly without an exception.
Just returning the words 'undefined' should be a good way to handle this.

• Include unit test that demonstrates that an equation containing the symbol Pi can be evaluated

For example: sixty six - 6 is solved correctly but sixty-six - 6 is not.

Previously reported as gunthercox/ChatterBot#1016