It explains that the part of the code that actually writes the data file to be used with NumPy's memmap was ran only once to ensure there was absolutely no cache in the Python side when running cold cache.

- can run write and/or read tests
- can choose between zblk formats
- can choose between all read test tools

- now there are parameters to run our tests with: wendelin+pandas, wendelin+numpy, numpy in memory, and numpy with memmaps
- fixed the selection of the ZBlkFormat
- test moved to performance_test_server.py (performance_test_client to be added next)

Conflicts:
	performance_test.py

- array is now populated by in-memory size and the number of rows is calculated automatically;
- array is populated using chunks of data to save some memory;
- parameter to choose between format ZBlk0 (one zodb object) and ZBlk1 (group of zodb objects).

It should be possible to chose between a normalized and a not normalized schema in the future.

- new context manager to measure code's performance
- now creates and fill arrays by memory size and not number of rows anymore
- only 1 row from mariadb is broadcasted to the entire array (doesnt spend a lot of time waiting mysql anymore)
- removed useless profile functions (just from this repository)
- added docstrings
- added pure wendelin.core/numpy sum test along with pandas

It should be possible to chose between a normalized and a not normalized schema in the future.

- new context manager to measure code's performance
- now creates and fill arrays by memory size and not number of rows anymore
- only 1 row from mariadb is broadcasted to the entire array (doesnt spend a lot of time waiting mysql anymore)
- removed useless profile functions (just from this repository)
- added docstrings

- added the disk size of the array object in zodb
- fixed profiling methods
- created a context manager to help time pieces of code
- timing numpy.sum() along with pandas.DataFrame.sum()