Path: utzoo!attcan!uunet!bionet!NCBI.NLM.NIH.GOV!pkarp
From: pkarp@NCBI.NLM.NIH.GOV (Peter Karp)
Newsgroups: bionet.molbio.genome-program
Subject: Re: feature table parsers - what's it all mean?
Message-ID: <CMM.0.88.654962493.pkarp@tech.nlm.nih.gov>
Date: 3 Oct 90 14:01:33 GMT
Sender: daemon@genbank.bio.net
Lines: 47

Let me answer the question in a different way.  From a computer
science point of view the GenBank features table is a language: every
possible feature is a sentence (string) constructed from an alphabet
of symbols (characters).  A grammar is a formal mechanism for
describing languages in a very precise way: a grammar specifies very
clearly what sentences are legal within that language and what
sentences are not.  Thus, by writing a grammar for the GenBank feature
table, GenBank is telling us what the allowable syntax of features is
-- without it we can only guess as to what strings we (and our
programs) can expect to see in feature tables.

A grammar is only a specification of a language -- it is not a
computer program for recognizing the language.  A parser is such a
program.  It takes as input a sentence in the language, and breaks the
sentence down into its constituent parts (for example, a parser of
English would break an English sentence down into verbs, nouns,
adjectives, etc.).  Parsers typically take different sorts of actions
when they see different grammatical elements (an English parser would
take one sort of action when it sees a verb and a different action
when it sees a noun).  Usually one cannot writer a parser without
referring to a grammar of the language.  A parser generator is a
program that automatically creates a parser program from a grammar
(the Unix YACC program is a parser generator).

The message here, as any computer scientist would have told you ten or
fifteen years ago, is that anyone who defines a language (such as a
text file that holds a database) that is to be used by a large number
of people, will make those people's lives MUCH easier if they define a
grammar for that language and make that grammar publicly available,
preferably in a format that can be used by a publicly-available
parser-generator program.  Without such a grammar people can only
guess as to the syntax of the language, people certainly cannot
generate parser programs automatically, and even writing parser
programs by hand can be VERY difficult.  Further, the very act of
writing a grammar makes the authors of such a language think very hard
about its properties: some languages are much easier than others for
computers to parse, and often the act of writing a grammar will cause
the language authors to transform the language into a new language
that is easier to parse.  By analogy, releasing a language without a
grammar is like having a society whose legal system operates without
written laws -- there are a lot of grey areas that will make many
people very nervous.

In summary, IF you release a computer database as a text file, PLEASE
consider writing a grammar for the language that you are using.

End of sermon.