How to make a calculator with strings and numbers as mixed input using parser python ply

Question

I would like to ask for help for an exercise to do a calculator which recognizes the English words and numbers in Python but now using PLY (Python Lex-Yacc)

The numbers and the operators can be given in two forms written as a string using English words, "plus" = "+", "two" = 2, "hundred twelve" = 112, etc.

An example could be these entries:

"twenty five divided by 5" or "25 / 5" or "twenty five divided by five"

the result should be the same, a number 5 (not a string).

" -3 times 4" will give -12

Division by 0 will give "Error" " 34 divided by 0" will give "Error"

This should work for several basic operators "-","+","x" and "/" (minus, plus, times and divided by) either if I type the mathematical symbols or I type in text or mixed.

Here are some parts of my code:

# ------- Calculator tokenizing rules

tokens = (
    'NAME','NUMBER', 'times', 'divided_by', 'plus', 'minus'
)

literals = ['=','+','-','*','/', '(',')']

t_ignore = " \t"

t_plus    = r'\+'
t_minus   = r'-'
t_times   = r'\*'
t_divided_by  = r'/'
t_NAME    = r'[a-zA-Z_][a-zA-Z0-9_]*'

  precedence = (
    ('left','+','-'),
    ('left','plus','minus'),
    ('left','times','divided_by'),
    ('left','*','/'),
    ('right','UMINUS'),
)

#Changed here the assigment def p_statement_assign(p): 'statement : expression times divided_by plus minus expression' variables[p[1]] = p[3] p[0] = None

def p_statement_expr(p):
    'statement : expression'
    p[0] = p[1]

def p_expression_binop(p):
    '''expression : expression '+' expression
                  | expression 'plus' expression
                  | expression '-' expression
                  | expression 'minus' expression
                  | expression '*' expression
                  | expression 'times' expression
                  | expression 'divided_by' expression
                  | expression '/' expression'''
    if p[2] ==   '+'  : p[0] = p[1] + p[3]
    elif p[2] == '-': p[0] = p[1] - p[3]
    elif p[2] == '*': p[0] = p[1] * p[3]
    elif p[2] == '/': p[0] = p[1] / p[3]

Are my tokens having a bad definition? How I can tell that the number can be introduce in English letter or with numbers?

The expression (p[2] == '+' : p[0] = p[1] + p[3]) has to have a single character. Why is not valid to write in this form p[2] == 'plus' : p[0] = p[1] + p[3] ?

---

I have added the code suggested by sfk, but I have still the problem to recognize the numbers and operators entered as text, in english words.

Generating LALR tables
WARNING: 12 shift/reduce conflicts
 Enter your input: calc > one + two
Undefined name 'one'
Undefined name 'two'
P1 is :  0
 Enter your input: calc > 1+2
P1 is :  3
3
 Enter your input: calc > 1 plus 2
Syntax error at 'plus'
P1 is :  2
2

Do you have any idea about what I am doing wrong?

Answer 1

First, add token definition for english words

t_plustext    = r'plus'

Add those new tokens to tokens

tokens = (
    'NAME','NUMBER', 'times', 'divided_by', 'plus', 'minus', 'plustext', ....
)

Finally, use those new token in you grammar this way :

def p_expression_binop(p):
    '''expression : expression '+' expression
                  | expression plustext expression
    '''

UPDATE : here is a working subset of the grammar

#!/usr/bin/python

from __future__ import print_function

import sys
import ply.lex as lex
import ply.yacc as yacc

# ------- Calculator tokenizing rules

tokens = (
    'NUMBER', 'times', 'divided_by', 'plus', 'minus', 'plustext',
    'one', 'two', 'three',
)

literals = ['=','+','-','*','/', '(',')']

t_ignore = " \t\n"

t_plustext    = r'plus'
t_plus    = r'\+'
t_minus   = r'-'
t_times   = r'\*'
t_divided_by  = r'/'
t_one = 'one'
t_two = 'two'
t_three = 'three'

def t_NUMBER(t):
    r'\d+'
    try:
        t.value = int(t.value)
    except ValueError:
        print("Integer value too large %d", t.value)
        t.value = 0
    return t

precedence = (
    ('left','+','-','plustext'),
    ('left','times','divided_by'),
    ('left','*','/'),
)


def p_statement_expr(p):
    'statement : expression'
    p[0] = p[1]
    print(p[1])

def p_expression_binop(p):
    '''expression : expression '+' expression
                  | expression plustext expression
                  | expression '-' expression
                  | expression '*' expression
                  | expression '/' expression'''
    if p[2] ==   '+'  : p[0] = p[1] + p[3]
    elif p[2] == '-': p[0] = p[1] - p[3]
    elif p[2] == '*': p[0] = p[1] * p[3]
    elif p[2] == '/': p[0] = p[1] / p[3]
    elif p[2] == 'plus': p[0] = p[1] + p[3]

def p_statement_lit(p):
    '''expression : NUMBER
          | TXTNUMBER
    '''
    p[0] = p[1]

def p_txtnumber(p):
    '''TXTNUMBER : one
         | two
         | three
    '''
    p[0] = w2n(p[1])

def w2n(s):
    if s == 'one': return 1
    elif s == 'two': return 2
    elif s == 'three': return 3
    assert(False)
    # See http://stackoverflow.com/questions/493174/is-there-a-way-to-convert-number-words-to-integers-python for a complete implementation

def process(data):
    lex.lex()
        yacc.yacc()
        #yacc.parse(data, debug=1, tracking=True)
        yacc.parse(data)

if __name__ == "__main__":
        data = open(sys.argv[1]).read()
        process(data)