On Fri, Nov 8, 2013 at 11:21 AM, Christopher Patton <
notifications@github.com> wrote:

"import qraat" gives you the following names: qraat.det, qraat.est,
qraat.rmg.{detect, params, blocks, run}, qraat.csv, and qraat.gps. We have
one instance of a true circular reference: qraat.rmg.params uses the class
qraat.csv. Since qraat.rmg is kind of a sub-module of qraat, we have a
cycle in our import tree. When you do "import qraat", Python knows where to
look as long as it's installed; if the code were not to be installed, this
would break.

From the current master branch, "import qraat" gives you:
qraat.csv - class not the module (alias for qraat.csv.csv)
qraat.det - class not the module
qraat.enum - function
qraat.gps - class not the module
qraat.pretty_printer - function
qraat.pulse_swig - module
qraat.qraat - recursive reference to itself

To get the rmg submodule you need to explicitly import it using "import
qraat.rmg". This is the way we want it so that's fine.

Note: It's impossible to import the qraat.{est, csv, gps, det} modules from
outside the scope of the qraat package because the qraat.init
overwrites the module reference with the class reference. This isn't
necessarily a problem, it's just kind of weird. We may want to rename the
modules something different from the classes so this is more explicit.

qraat.csv and qraat.rmg.params aren't circular references because qraat.csv
doesn't import qraat.rmg or any of it's submodules. However qraat.csv and
qraat.est are circular because qraat.csv imports qraat which imports
qraat.est which imports qraat which imports qraat.csv.

The proper way to fix this would be to eliminate any circular
dependencies. We could do this by making "rmg" and "qraat" separate modules
or by moving moving qraat.csv to qraat.rmg.csv. I don't think the changes
you made would do it.

The changes I made eliminate both the recursive reference qraat.qraat and
the two "circular" references (qraat.csv - qraat.rmg.params and qraat.csv -
qraat.est) by eliminating the unneeded import of qraat in qraat.csv and
changing the import * in qraat.init (generally bad in python). It's
okay for qraat.rmg.params to import qraat.csv as long as qraat.csv doesn't
try to import qraat.rmg.params, and thus complete the circle (same with
qraat.est).

WRT catching "import MySQLdb", this was necessary to import the module
on the nodes, which don't have python-mysqldb installed. We could
selectively disable features which won't run on a particular platform, but
in my opinion, there's not reason to worry about this at this stage.

That makes sense. It might make sense to throw a warning that MySQLdb
isn't imported so when you get an "mdb doesn't exist" error it will be more
easily tracked down.

Overall, I think we can put off this change. Circular dependencies are
bad, but it's not broken. I suggest we leave this as an open issue and come
to it later on in the future.

It's not broken in the sense that it works now but there are some really
weird things that can happen down the road if you aren't aware of these
intricacies.
-Todd

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/6#issuecomment-28089898
.

Circular dependance in Python module #6

Description

Metadata

Metadata

Assignees

Labels

Type

Fields

Projects

Milestone

Relationships

Development

Issue actions