You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
ATM all ArgOrdering subclasses return the argument as strings. In #1883 support for getting the same information as PSyIR was added, but the old, string-based results are still supported and used everywhere (the PSyIR results are only needed for LFRic driver creation atm).
The string-handling should be replaced with the PSyIR based results.
The text was updated successfully, but these errors were encountered:
Related to this, upon construction, an instance of ArgOrdering stores a reference to a SymbolTable. As noted by @sergisiso (#2313 (comment)) this is dangerous as we end up with a double reference which won't work e.g. if the tree is copied.
Whilst ArgOrdering makes the LFRic world go round, there is a new base class implementation based on LFRic metadata in MetadataToArgumentsRules.py. My idea is to use this to replace ArgOrdering with this and then modify/re-write the subclasses as needed. There is PR #2079 that is that is looking to fully implement the equivalent of kernel_interface as a first step (although it is already used in PSyAD for index mapping).
ATM all ArgOrdering subclasses return the argument as strings. In #1883 support for getting the same information as PSyIR was added, but the old, string-based results are still supported and used everywhere (the PSyIR results are only needed for LFRic driver creation atm).
The string-handling should be replaced with the PSyIR based results.
The text was updated successfully, but these errors were encountered: