hebi: new simili-BI module

[resteve, 2013-06-07 14:44:09]

https://codereview.appspot.com/9943044/diff/1/setup.py
File setup.py (right):

https://codereview.appspot.com/9943044/diff/1/setup.py#newcode14
setup.py:14: description='Library to access Tryton server as a client',
in README I read: A library to create pre-computed tables

[ced, 2013-07-08 15:10:17]

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py
File hebi/__init__.py (right):

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode14
hebi/__init__.py:14: class ScalarColumn(sql.Column):
Could be sql.Literal

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode32
hebi/__init__.py:32: class flat(Column):
Flat

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode36
hebi/__init__.py:36: class relation(flat):
Relation

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode43
hebi/__init__.py:43: class tree(Column):
Tree

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode87
hebi/__init__.py:87: class leaf(Column):
Leaf

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode95
hebi/__init__.py:95: def __init__(self, prefix='', *columns):
I would separate grouping columns and leaf because leaf should always be at the
end.
So having this signature sounds less error prone:

Schema(grouping=(), aggregate=(), prefix='')

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode106
hebi/__init__.py:106: "{} should be a instance of Column".format(name))
an instance

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode107
hebi/__init__.py:107: self.cumulate = all(issubclass(l.operator,
CUMULATABLE_AGGREGATORS)
It should be better to have it as a property in case leafs is modified after the
instantiation.

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode110
hebi/__init__.py:110: def groupbys(self):
groupbys -> groupings

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode113
hebi/__init__.py:113: while level < len(keys):
for i, key in enumerate(keys):
   yield keys[:len(keys) - i]

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode127
hebi/__init__.py:127: def explode_table(self, table, cursor):
delete should always be run before any processing.

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode132
hebi/__init__.py:132: if isinstance(self.branches[groupby[-1]], tree):
I'm wondering if not tree could not be a simple case of tree?

https://codereview.appspot.com/9943044/diff/1/hebi/__init__.py#newcode183
hebi/__init__.py:183: for col_name in groupby + self.leafs.keys()]
This construction is based on the specific way of constructing
create_groupby_query.
I find having both in separate methods dangerous for maintenance  to keep
consistency.

[ced, 2013-09-25 15:09:17]

https://codereview.appspot.com/9943044/diff/15001/hebi/__init__.py
File hebi/__init__.py (right):

https://codereview.appspot.com/9943044/diff/15001/hebi/__init__.py#newcode161
hebi/__init__.py:161: print parents
print statement

https://codereview.appspot.com/9943044/diff/15001/hebi/__init__.py#newcode184
hebi/__init__.py:184: print '+' * 80
print stamenet

https://codereview.appspot.com/9943044/diff/15001/hebi/__init__.py#newcode189
hebi/__init__.py:189: print tuple(update_qry)
print stament

[vvanderleun, 2013-10-23 16:36:56]

Right now Hebi is not Python 2.6 compatible, while this is stated in setup.py.

https://codereview.appspot.com/9943044/diff/47001/hebi/__init__.py
File hebi/__init__.py (right):

https://codereview.appspot.com/9943044/diff/47001/hebi/__init__.py#newcode97
hebi/__init__.py:97: self.leafs = collections.OrderedDict()
According to setup.py Hebi is supposed to be Python 2.6 compatible. OrderedDict
is not in collections in 2.6, but there's a drop-in replacement package
available [1]

I always patch hebi by adding "from collections OrderedDict" at the top unless
an ImportError is raised, and then try "from ordereddict import OrderedDict".
And I change all "collection.OrderedDict" references to just "OrderedDict".

[1] https://pypi.python.org/pypi/ordereddict

[ced, 2014-08-05 11:47:38]

I think the tree column process could be replaced by a recursive SQL query that
flatten the tree by duplicating each leaf for all his grand-parents.
With this kind of query the code of explode_table could be unified and simplier.