fix allowed table and field names about pydal HOT 16 CLOSED

Problem could be back compatibility

from pydal.

it was that way. Then cleanup was relaxed to.... pretty anything.
Versions slipped by, and now we face a problem.

from pydal.

I agree, and we should have the same approach for situations such as:

db = DAL("sqlite:memory")
db.define_table('a', Field('1'))

raises

AttributeError: 'Field' object has no attribute 'iteritems'

while

db = DAL("sqlite:memory")
db.define_table('a', Field(1))

raises

SyntaxError: Field: invalid field name: 1, use rname for "funny" names

In addition, the following works at 'pydal' level:

db = DAL("sqlite:memory", migrate=False)
db.define_table("1", Field('a'))

but then it fails at 'db' level:

db = DAL("sqlite:memory")
db.define_table("1", Field('a'))

raises

sqlite3.OperationalError: near "1": syntax error

from pydal.

this should be fixed before the next release to avoid finding ourselves in 6 months locked-in for backward compatibility

from pydal.

I've moved this to 15.04 milestone. If anybody is interested into fixing this, let us know.

from pydal.

this is quite an easy fix, once we all agree. There's nothing documented in the book, so a clear and large statement in the CHANGELOG would suffice.

I didn't read anything conclusive on web2py-developers, but I strongly think that this should be closed once and for all.
I (re)vote for cleanup() restricting table and field names to valid python identifiers. If I'm not mistaken, this is the reference.
Pinging everybody for the call-to-arms: @mdipierro , @gi0baro , @michele-comitini , @abastardi , @ilvalle , @leonelcamara ... feel free to chime in.

from pydal.

+1 on @niphlod vote

from pydal.

ok, I think we can safely assume the only responsible for whining users would be me and @gi0baro . I'm ready to take the heat.

@gi0baro : ^[^\d\W]\w*\Z should be the regex.
BTW: I'm not sure about leaving up the whole python3 range for identifiers, which accounts for unicode characters as well. I'd stay with valid python2 identifiers.

from pydal.

@niphlod ok with that

from pydal.

ok, will send a PR soon enough.

from pydal.

@gi0baro : another question... I'd go for the same routine for both tablenames and fieldnames, which is basically:

• a str
• a valid python identifier
• not a python keyword
• not starting with underscore
• not containing dots

table would have also the restriction of not being a yet-existing attribute of DAL.

tl;dr: adding "a valid python identifier" constraint to the existing one on Table and all of the above for Field

from pydal.

umpf. quite there. How the heck should we check that è is not a str in python3 ? Meaning... we don't check for Field and Tablenames in python2 to be valid basestr that would make things easier porting to python3... we only check with isinstance(obj, str) . How can we make a validation on python3 for the same thing ?

from pydal.

@niphlod sorry, I was AFK. Anyway, I'm missing your point, since

isinstance('è', str)

is True both on python 2 and python 3. And on python 2 isinstance('è', basestring) is True as well.
If we want to remove special characters I think we should use a regex.

from pydal.

right ........ will do this evening.

from pydal.

sorry if I chime in late

@niphlod, does this help?
https://stackoverflow.com/questions/5474008/regular-expression-to-confirm-whether-a-string-is-a-valid-identifier-in-python

2015-05-20 10:40 GMT+02:00 niphlod [email protected]:

right ........ will do this evening.

—
Reply to this email directly or view it on GitHub
#67 (comment).

from pydal.

unfortunately not. in py3 it would allow also unicode, since \W is in. Seeing the code, I won't touch "cleanup()" as it seems that the validation is both in Table and Field __init__
I'm in a place where no python exist but I think I'll refactor a bit those checks and just leave ^[^\d_][_0-9a-zA-Z]+$

#current
if (not isinstance(tablename, str) or tablename[0] == '_'
            or hasattr(DAL, tablename) or '.' in tablename
            or REGEX_PYTHON_KEYWORDS.match(tablename)
            ):
#will probably become
# in regex.py
NEWREGEX = re.compile('^[^\d_][_0-9a-zA-Z]+$')
....
if (not isinstance(tablename, basestr) or hasattr(DAL, tablename)
            or NEWREGEX.match(tablename)
            or REGEX_PYTHON_KEYWORDS.match(tablename)
            ):

from pydal.