How hashing works in xolotl and cpplab
This document describes how xolotl and cpplab use hashing to verify the identity of models and to select the correct binary to use.
What is a hash?¶
A hash is a short string that is used to uniquely identify a arbitrary-length piece of data. A hash is created by a hash function that accepts data of any length, and returns a string of a fixed length. It may seem impossible that any data of any length can be uniquely identified by a short string; and this is true. In practice, however, a good hash function avoids collisions of these sorts, and most data resolves into unique hashes.
xolotl uses hashes to identify models. Since models can be nested and hierarchical and sometimes quite large and complex, a hash is a convenient way or referring to a model.
How models are hashed¶
The structure of the xolotl object is what is hashed. This means that the combination of network components (e.g. compartments, conductances, etc.) with their names and connections is assigned a cryptographic string of alphanumeric characters. We use the well-studied MD5 algorithm. Crucially, parameters of the model are not hashed. This means that families of models, those with the same structure/equations but different parameter values have the same hash.
When a change is performed to the xolotl object in MATLAB, a new compiled C++ binary is created only if the hash changes (i.e. the structure of the model changed). This means that running models that only differ by parameters is extremely fast and easy.
integrate method is called, the hash of the xolotl object is compared
with the hash of the compiled binaries and the correct one is chosen. The xolotl
serialized to get a list of parameters, which are passed as function arguments.
This way, you only need one compiled binary per unique xolotl structure. Parameters can