updated design of the block access layer

clifordjoshy commented 1 year ago

The designs of this layer was unnecessarily verbose, especially for renameRelation and renameAttribute which is mostly just code. I have rewritten them to be better representative of the core algorithm. The new designs and the specific changes are mentioned below Additionally, the project method has been added and reset has been handled in the block access layer. these function calls have been updated throughout

linearSearch

minor corrections and additions
replaced a switch statement with the same case body throughout with an if statement checking for all the required equalities and conditions
updated the comments to reflect the above change
removed the PRJCT operator and moved that functionality to project function

search

removed the RST and PRJCT operators. reset is now handled in RelCacheTable::resetSearchIndex and project in the project method.

renameAttribute

reworded the docs and replaced a lot of code with the necessary descriptions

deleteRelation

rewrote the docs to be less verbose
for deleting the record blocks, the existing design would call releaseBlock even if there were no record blocks for the relation. fixed that.
in the above case, for seeing if we were at the last record block the old design would allow checking for currBlock == lastBlock OR nextBlock == -1. the first case would have been an infinite loop in the case of no record blocks.

project

the project operation is now handled here. all functions have been updated to match this schema.

clifordjoshy commented 1 year ago

TODO; commit the changes in the search function to the main repo before merging

clifordjoshy commented 1 year ago

Additional Suggestion

Instead of manually setting the search index to {-1, -1} before every linearSearch using the RelCacheTable, maybe we could create a method to reset the search index? it will make the code a lot more readable. And additionally, would also remove the need for the RST operator in BlockAccess::search(). However, it is not a core functionality of the layer. So, I'm not sure if it's a valid change?

clifordjoshy commented 1 year ago

Additional Suggestion

Instead of manually setting the search index to {-1, -1} before every linearSearch using the RelCacheTable, maybe we could create a method to reset the search index? it will make the code a lot more readable. And additionally, would also remove the need for the RST operator in BlockAccess::search(). However, it is not a core functionality of the layer. So, I'm not sure if it's a valid change?

has been handled in the last commit. rst and project have been moved to separate functions

NITCbase / nitcbase.github.io