feat: inner join #71
feat: inner join #71
Conversation
… AssertSortedChip
| /// Multiplicity of idx in output_table | ||
| pub out_mult: T, | ||
| /// Indiates if this row is extra and should be ignored | ||
| pub is_extra: T, |
There was a problem hiding this comment.
The behavior of is_extra is almost same to is_alloc. Can we reuse that logic?
There was a problem hiding this comment.
Reuse that logic where?
| impl MyFinalTableAir { | ||
| #[allow(clippy::too_many_arguments)] | ||
| pub fn new( | ||
| t1_output_bus_index: usize, |
There was a problem hiding this comment.
Maybe reuse the parameter struct I mentioned in controller constructor
There was a problem hiding this comment.
Here we only use 3 bus indices
|
|
||
| #[allow(dead_code)] | ||
| struct TableCommitments<SC: StarkGenericConfig> { | ||
| t1_commitment: Com<SC>, |
There was a problem hiding this comment.
It's a bit hard to know what t1/t2 refers. Can we call its formal names? From chatgpt:
-
Parent Table (or Referenced Table): This is the table that contains the primary key, which is referenced by the foreign key in the other table. It is often referred to as the "parent" because it holds the original values that are being referenced.
-
Child Table (or Referencing Table): This is the table that contains the foreign key, which references the primary key in the parent table. It is often called the "child" because it depends on the parent table for its foreign key values.
There was a problem hiding this comment.
Do you only the the commitments struct or everywhere?
I think it's annoying and verbose to use those names everywhere. For example, I think parent_table_intersector_bus_index becomes too much. I added a comment above the constructor to note which table is T1 and which is T2.
| where | ||
| Val<SC>: Field, | ||
| { | ||
| Self { |
There was a problem hiding this comment.
Assert validity of
fkey_start: usize,
fkey_end: usize,
t1_idx_len: usize,
t1_data_len: usize,
t2_idx_len: usize,
t2_data_len: usize,
idx_limb_bits: usize,
decomp: usize,
There was a problem hiding this comment.
what is there to validate for idx_len, data_len, limb_bits, decomp?
|
@nyunyunyunyu can you take a look at this again |
* feat: sorted_limbs chip checking each limb less than limb_bits bits * wip * feat: completed sorted_limbs chip with tests * wip * feat: SortedLimbsChip with LessThan subchip * feat: less_than subchip refactored * feat: rename SortedLimbsChip to AssertSortedChip and write LessThanChip tests * chore: change name of assert sorted chip * chore: fix names in tests for AssertSortedChip * chore: address comments * chore: cleanup * chore: change MAX from generic to instance field for LessThanChip and AssertSortedChip * wip: added extra bits to middle chip trace * wip * adding connection to IsEqualVec chip * feat: IsLessThanChip to compare two numbers * feat: IsLessThanTuple subchip for different limb_bits * test: added tests for partially and non-allocated pages * test: added negative tests * added constraints * adding comments * feat: IsLessThanTupleChip subchip in AssertSortedChip * renaming * removing TODO comments * fixing clippy * chore: renaming to idx and data * chore: address comments first pass * chore: moving page_controller inside page_read * chore: refactor AssertSorted, IsEqual, IsLessThan, and IsLessThanTuple chips * chore: address comments * chore: eliminate high dim poly from IsLessThanTupleChip * chore: fix tests * chore: address comments for AssertSortedChip * chore: cleanup AssertSorted * chore: cleanup * chore: include roundtrip flatten and from_slice tests * feat: flatten and from_slice for IO and Aux columns * wip: final page chip * remove txt file * feat: final_page_chip * feat: integrating all chips for page_read_write * fixing lib * using field bits() * adding extra communication between checker and final chip * muting clippy * renaming to AIR * adding comments * comment fix * optimization to one less bus * comments * moving old page_controller inside page_read * updateing comment * wip: inner join * feat: more general FinalPageAir * making IsEqualVecAuxCols clonable * enforcing unallocated rows to be zero * aligning with new less than chip * wip * using Air::eval instead of SubAir::eval * fix comment * wip * removed arguments from load_page for readability * addressing comments and adding common/ directory with Page struct * wip * controller complete * controller complete * adding less than constraints * adding comments * more comments * added keygen and prove functions to controller * added verify function to controller * addressing comments * merging new chanegs * some renaming * addressing comments * removing too_many_arguments bypassing --------- Co-authored-by: bfan <76703988+bfan05@users.noreply.github.com> Co-authored-by: Jonathan Wang <31040440+jonathanpwang@users.noreply.github.com>
Straightforward implementation from the Inner Join spec: https://docs.google.com/document/d/1HudyqvMzlWKA4eT0IjkIKCbidQCH3utSXXR-9vm_xcI/edit
Also added some helper function to the Page struct that are useful in testing.
Closes INT-1532