rizza is now using the field type to create an #29
Conversation
| self.base_dir = Path.home().joinpath('rizza') | ||
| self.base_dir = Path(os.getcwd()) |
There was a problem hiding this comment.
Rizza is meant to operate out of a user's home directory, not current directory
There was a problem hiding this comment.
I did this just to set up my project. With this upgrade, it worked for me outside the home's directory.
| 'MACAddressField': 'gen_mac', | ||
| 'NetmaskField': 'gen_netmask', | ||
| 'StringField': 'gen_string', | ||
| 'URLField': 'gen_url', |
There was a problem hiding this comment.
mapping fields like this may stepping over the line that is providing too much help to Rizza instead of letting Rizza figure things out with experimentation
There was a problem hiding this comment.
I disagree, because what we want to achieve in first step is fast genetic programming, is always easier to make it more complex without a hints.
A) Now we are talking about 15 different types, 2 are using other entities, 2 are more complex types lst and dict.
And now they are exactly match. So complexity is only in choosing the right fields and arguments, not matching the types instead.
B) In your previous solution you used 37 types,
list(entity_tester.EntityTester.pull_input_methods() without float list and dict which are actually used in the nailgun.
so each field and attribute has to firstly guess the type which is complexity of number_of_fields^field_types + number_of_args^field. And we didn't even start to match the fields and arguments required for the entities.
A broader search of space in option B can found results in BZ creation, but it would take forever. You might run it separately, but anyway you firstly need a way to look at the answer and say okay this is fine/wrong. Doing it manually might be overkill.
| @@ -34,6 +34,7 @@ def _breed_pair(self, gene_list1, gene_list2): | |||
| Random mutation is then given a chance to change the child. | |||
|
|
|||
| """ | |||
| # TODO now with nested lists and dict this requires rework | |||
There was a problem hiding this comment.
this is now the bottleneck which requires additional work. As mutation in some cases does not work the program will fail, but it shouldn't be a lot of work to fix it.
| self.config | ||
| )) | ||
| return lst | ||
| # TODO dict |
There was a problem hiding this comment.
the same needs to be done for dict,
and of course dict of lists and lists of dict.
| self.load_config() | ||
| self.RIZZA['CONFILE'] = self.cfg_file |
There was a problem hiding this comment.
I also played a bit with clean config config rizza --clear, to reload environment vars into the conf file and therefore to code.
| # we want to always use current directory as base for tests | ||
| if 'tests' in str(self.cfg_file): | ||
| self.cfg_file = Path().joinpath(self.cfg_file) | ||
| elif self.cfg_file != str(Path(self.cfg_file).absolute()): | ||
| self.cfg_file = self.base_dir.joinpath(self.cfg_file) | ||
| self.RIZZA['CONFILE'] = self.cfg_file | ||
| self.cfg_file = str(self.base_dir.joinpath(self.cfg_file)) |
There was a problem hiding this comment.
it wrote !!python-bla-bla-bla into the configuration file so it was required to retype to string
|
I believe that it is successfull attempt, just not finished |
this is a draft
Please read the comments and todo's.
I search for all usages in nailgun entities file and all of them are created mainly just by the simple fauxfactory method.
With this knowledge, I map field type to fauxfactory method, so it can be created.
There are still some waiting for implementation, but this is the first step.
Also. I disabled OneToOneField and OneToManyFields in case there are no dependencies allowed.
These changes gradually improve the speed of search.