We have a data import / transformation tool currently written in C++ which is processing mostly textual (e.g. timestamps in several different formats) data and applies some normalization / sanitization.
The current solution is very fast and performance is critical. E.g. we optimized to minimize / avoid object allocation in the parsing loop as this gets called for each row we process and can have a serious performance impact.
What we want to achieve is being flexible with adding new input formats or applying additional transformations simply by changing a script/configuration file instead of recompiling the application.
The question is: would we be able to achieve a comparable performance using the Python Pandas data processing framework, as most of it also seems to use low-level optimized C code.
("comparable" here means performance difference is within a margin which may be compensated by adding 2-4 additional threads)
Is Pandas the right tool for this job or are there different suggestions?