Ambiguous IDs in Kappa grammar

[Sandro]

Hi!

There is an issue with the Kappa grammar I encountered when writing an Emacs mode for Kappa. Here is what the KaSim manual says about IDs:

> Symbol Id can be any string generated by regular expression [a-zA-Z0-9][a-zA-Z0-9_-]*.

This is ambiguous when used in variable definitions as illustrated below:

    # These variable definitions illustrate the ambiguity between numerals

    # and IDs in the Kappa grammar.

    %var: '1A'      1A    # 1A   is an agent name

    %var: '1E'      1E    # 1E   is an agent name

    %var: '1E+1'    1E+1  # 1E+1 is a numeral

    %var: '200'     200   # 200  is a numeral (no agent signature)

    %var: '100'     100   # 1E0  is both a numeral and an agent name

    %var: '1E0'     1E0   # 1E0  is both a numeral and an agent name

In practice, KaSim doesn't accept agent signatures with "numeric" agent names. E.g. the following code

    # Numeric IDs in agent signatures

    %agent: 1A(x)   # Declaration of agent 1A      => works.

    %agent: 1E(x)   # Declaration of agent 1E      => works.

    %agent: 100(x)  # Declaration of agent 100     => fails!

    %agent: 1E0(x)  # Declaration of agent 1E0     => fails!

will generate a parser error:

    Error (id_bug.ka) line 19, character 11: Malformed agent

    signature, I was expecting something of the form '%agent:

    A(x,y~u~v,z)'

Is there any specific reason for allowing IDs with an initial digit?

Cheers

/Sandro

[Jean Krivine]

Yes some proteins are named like that. I don't remember which though. But they are the reason for allowing names to start with a digit

[Vincent Danos]

14-3-3

[Bill Hlavacek]

The gene name for a 14-3-3 protein would not be # but rather plain text.

--Bill

[Jean Krivine]

What do you mean Bill?

[Bill Hlavacek]

Rather than name the agent for the protein 14-3-3alpha/beta, you could name is YWHAB, which is the gene name for this protein. The gene names are computer friendly.

[Sandro Stucki]

Sorry for re-posting. Forgot to include the mailing list in the reply address.

---------- Original message ----------
From: Sandro Stucki <sandro...@gmail.com>
Date: Mon, May 7, 2012 at 1:11 PM
Subject: Re: Ambiguous IDs in Kappa grammar
To: Bill Hlavacek <wshla...@gmail.com>


Are there a lot of people using this "feature"? It seems cleaner to me
to have an unambiguous grammar. The most common solution, I guess, is
to allow only alphabetical characters and "_" as the first character
in an identifier. Unless, of course, that would break a lot of
existing models. Then again,

 sed -e 's/[0-9][0-9]*-[0-9][0-9]*-[0-9][0-9]*/_&/g' oldfile > newfile

would probably do the trick.

What do you think?

/Sandro

[Jean Krivine]

I am happy with it.

[Donal Stewart]

Happy with the existing grammar or happy with the suggested modification ?

Meta-ambiguity ;)

> <jean.k...@gmail.com <mailto:jean.k...@gmail.com>>

[Jean Krivine]

Haha no happy with the proposed modification of the grammar.