Um... extensible parsing?

[holger krekel]

Philip Swartzleonard wrote:
> 
> I am interested in making a program in python that helps people edit the 
> gameworld definition files used in the Angband and it's many and various 
> varients. They are a fairly simple text file format, and each record 
> follow a mostly set pattern. For exapmle, here's the format for one of 
> the files from ToME:

ask any questions you need to about the following code
which should help you get started.

# paste this to your interpreter and
# have a /tmp/datafile ready
specs=[]
for line in open('/tmp/datafile').readlines():
    if not line or line.startswith('N'): 
        specs.append({})
    name,rest=line[:1],line[2:]
    if name:
        sep = '|' in rest and '|' or ':'
        vals=map(lambda x: x.strip(), rest.split(sep))
        specs[-1].setdefault(name, []).extend(vals)

# then enter e.g.

>>> specs[0]
{'B': ['CLAW', 'HURT', '1d2', 'CLAW', 'HURT', '1d2', 'BITE', 'HURT',
'1d3'],
 'D': ["A woman's face on the body of a vicious black bird."],
 'E': ['0', '1', '1', '0', '1', '0'],
 'F': ['FEMALE',
       'CAN_FLY',
       'WILD_TOO',
       'WILD_MOUNTAIN',
       'DROP_CORPSE',
       '',
       'RAND_25',
       'ANIMAL',
       'EVIL',
       'MORTAL',
       'BASEANGBAND'],
 'G': ['H', 'D'],
 'I': ['120', '3d8', '16', '22', '10'],
 'N': ['157', 'Black harpy'],
 'O': ['0', '0', '0', '0'],
 'W': ['9', '1', '600', '19']}
>>>

> Anyway, I want to make a generic editor program that reads some datafile 
> to determine the structure of the 'info' file it needs to edit, and 
> constructs it's interface from there. I'd assume i'd have some kind of 
> list interface to deal with the records (using the N line information 
> that all of thie files that I want to edit use), and that I'd have 
> customized controls for editing the different types of data (e.g. 
> flag_checkboxes, string_field, dice_field, number_field). What i -don't- 
> have a clue how to do is the part that puts the pieces together 
> depending on the situation, and the parts that actually parse the 
> various files. What i'm asking for is any pointers or insight that could 
> help me solve this problem more efficently. Thank you.

- get used to working flexibly with dictionaries

- use a class basically like this: 

    class niceaccessor:
       def __init__(self,specdict):
         for attr,value in interpret(specdict):
            setattr(attr,value)
  
  so that you can do 
  print niceaccessor.name     
  if ANIMAL in niceaccessor.flags:
        do this or that
  ...

- use such a class to make up some frontend

- write a routine which takes any existing
  key/value of an niceaccessor-instance  
  and writes out a record.

- think about whether a string that can
  be converted to a number should always
  be converted or whether you have 
  to actually specify this.

hope that helps, 

    holger