maeparser is a parser for Schrodinger Maestro files.
Structure files (.mae,.maegz,.mae.gz) can contain multiple structures delimited by "f_m_ct". See MaeConstants.hpp for standard block and property names.
To read a structure,
#include "Reader.hpp"
...
FILE* f = fopen("test.mae", "r");
schrodinger::mae::Reader r(f);
std::shared_ptr<schrodinger::mae::Block> b;
while ((b = r.next(schrodinger::mae::CT_BLOCK)) != nullptr) {
// Parse structure
}
fclose(f);See also test/UsageDemo.cpp, which reads an example structure and stores it in a dummy Molecule class.
Why do we have a recursive descent parser for mae files instead of using a parser generator like ANTLR or lex/yacc? The main reasons are that the mae format language is 1) pretty simple, 2) unlikely to change significantly, and 3) not a context free grammar. In addition, speed of parsing these files is important.
In what way is the current version of the language not a CFG? Special tokens
like block opener { and key/value separator ::: can also be string
values because the quotes on string values are not required. This results in
complication and pain in attempts to define a grammar.
There are many molecule formats out there, and the significant strength of this one is that it exactly fits the use case of Schrödinger's physics-based modeling. As the primary (and only lossless) Schrödinger output format, any package that wishes to implement lossless data extraction of Maestro output needs to interact directly with this format. This is not an intentional limitation, but is due to the nature of chemical storage formats: it's extremely hard to get a format that's both 1) Flexible enough to hold any type of data and 2) Not so flexible that each user has to implement their own rules. The Maestro format avoids this paradox by having the exact flexibility that Schrödinger's physics based backends require, without additional flexibility that other use cases might demand.
In supporting Schrödinger's backend suite, maeparser is able to handle output from:
- Molecular Dynamics applications, such as Desmond and FEP+
- Ligand-Protein Docking applications, such as Glide
- Homology Modeling and folding applications, such as Prime
- Ligand-based search applications, such as Phase and Phase Shape
- Quantum Mechanics applications, such as Jaguar
- Protein-Protein Docking applications
- Many other backends used in both Life and Material Sciences