How to in Python

[MartinRinehart]

I've got a pointer to a position in a line of code that contains
either a digit or a period (decimal point). I've got this comment:

Numbers are one of these:
integers:
digit+
0xhex_digit+
decimals:
digit+.digit*[E['+'|'-']digit+]
.digit+[E['+'|'-']digit+]
digit+[.digit*]%
.digit+%

Common metacode: '*' = 0 or more, '+' = 1 or more, [] = optional, | =
or, ...

Now I need to instantiate the comment. How would an experienced Python
coder proceed?

 

[Gabriel Genellina]

I've got a pointer to a position in a line of code that contains
either a digit or a period (decimal point). I've got this comment:

Numbers are one of these:
integers:
digit+
0xhex_digit+
decimals:
digit+.digit*[E['+'|'-']digit+]
.digit+[E['+'|'-']digit+]
digit+[.digit*]%
.digit+%

Common metacode: '*' = 0 or more, '+' = 1 or more, [] = optional, | =
or, ...

Now I need to instantiate the comment. How would an experienced Python
coder proceed?

Do you have to validate input based on that grammar? That is, do you want
a function like this?

def is_valid_number(input):
if ....
return True
return False

Or, do you want to consume characters starting from your pointer, stopping
when an invalid character appears?

This looks like one of those few cases where "Use a regular expression"
may be a good answer. If you like a more Pythonic approach, try using
Pyparsing <http://pyparsing.wikispaces.com/>

 

[John Machin]

I've got a pointer to a position in a line of code that contains
either a digit or a period (decimal point). I've got this comment:

Numbers are one of these:
integers:
digit+
0xhex_digit+
decimals:
digit+.digit*[E['+'|'-']digit+]
.digit+[E['+'|'-']digit+]
digit+[.digit*]%
.digit+%

Common metacode: '*' = 0 or more, '+' = 1 or more, [] = optional, | =
or, ...

Now I need to instantiate the comment. How would an experienced Python
coder proceed?

Use a proper lexer written by somebody who knows what they are doing,
as has already been recommended to you.

 

[MartinRinehart]

Use a proper lexer written by somebody who knows what they are doing,
as has already been recommended to you.

My lexer returns a MALFORMED_NUMBER token on '0x' or '0x '. Try that
in Python.

 

[MartinRinehart]

Do you have to validate input based on that grammar?

I've built a standalone tokenizer. It returns an array of Token
objects. These include tokens such as UNCLOSED_QUOTE and
MALFORMED_NUMBER ('1E' not followed by sign or digit, for instance).

You could use this in a code editor to colorize. You could use it in a
documentation writer to emit source in HTML. You could use the tokens
in a parser.

 

[MartinRinehart]

If I get to add multi-line strings today, I'll have a complete
tokenizer. Interior looks a lot like C minus semi-colons. (Though I
did figure out that there wasn't any need for tokens that didn't come
from a real to have a doubleValue field. In C++ or Java all the Tokens
had a doubleValue, because one needed it.)

Theory: there are some problems which, by their nature, end up looking
a lot like C, regardless of language.

I wrote a Perl-to-HTML pretty-printer. It lets you embed HTML in
comments and creates a table of contents hyperlinked to the individual
functions. It's at http://www.MartinRinehart.com , output and source
code, in the Articles section.

I started with a Perl-style solution: regex for the chars left of "#"
and for the chars to the right. One line of Perl. Nice, except that it
won't work when applied to itself. It will split the line based on the
"#" in the regex, of course. Kludged around that, but then met "#"
embedded in strings, "#" embedded in regex passed to functions, ...

Ended up marching down the input line, one character at a time, like a
C program.

 

[Chris Mellon]

My lexer returns a MALFORMED_NUMBER token on '0x' or '0x '. Try that
in Python.

Is there some reason that you think Python is incapable of
implementing lexers that do this, just because Python lexer accepts
it? Note that if you're using your lexer to mark up or pretty print or
whatever Python source, it's wrong - 0x is (rightly or not) a valid
Python literal.

 

[MartinRinehart]

Is there some reason that you think Python is incapable of
implementing lexers that do this, just because Python lexer accepts
it?

Absolutely not. My opinion is that it's a bug. A very, very minor bug,
but still six-legged.

Note that if you're using your lexer to mark up or pretty print or
whatever Python source, it's wrong - 0x is (rightly or not) a valid
Python literal.

My lexer is for my language, Decaf, which, in this particular, is the
same as Python. Here's what I find at at python.org/ref: (2.4.4).

hexinteger ::= "0" ("x" | "X") hexdigit+

Implementation differs from specification. In this case, I think the
spec is more sensible.

 

[Chris Mellon]

Absolutely not. My opinion is that it's a bug. A very, very minor bug,
but still six-legged.

I understand the argument but I'm wondering why you're asking
questions on this list about it. You don't seem to be implementing the
lexer in Python (because otherwise the question wouldn't have come
up), and you don't seem to be parsing Python code. Where's the python
angle? Not that this list is totally intolerant of offtopic
discussions, but you don't seem to even have a hint of Python interest
here - the questions seem to be better suited for comp.parsers.general
or something.

My lexer is for my language, Decaf, which, in this particular, is the
same as Python. Here's what I find at at python.org/ref: (2.4.4).

hexinteger ::= "0" ("x" | "X") hexdigit+

Implementation differs from specification. In this case, I think the
spec is more sensible.

I tend to consider "what the parser actually accepts" rather than
"what the grammar specifies" to be normative when writing pretty
printers or syntax highlighters.

 

[MartinRinehart]

You don't seem to be implementing the
lexer in Python

I am absolutely implementing my language in Python, a language I have
now been writing for two entire weeks. This list has been more than
helpful, tolerating numerous newbie questions.