Implementation of intersection algorithm for regular grammars #34
Conversation
| curr_front = list() | ||
| visited = Vector.sparse(BOOL, num_verts_inter) | ||
|
|
||
| for regex_start_st in regex_start_states: |
There was a problem hiding this comment.
Can we create a multiple-source version instead of multiple running of single-source version?
|
|
||
| for symbol in regex: | ||
| if symbol in graph: | ||
| with semiring.LOR_LAND_BOOL: |
There was a problem hiding this comment.
I guess, ANY_PAIR should be used here.
|
|
||
| visited[prev_vert_index] = True | ||
|
|
||
| if not found.iseq(vect) and len(list(found.I)) == 2: |
There was a problem hiding this comment.
Looks too complicated. Can it be expressed in terms of GraphBLAS vector operations?
There was a problem hiding this comment.
What if we have multiple ways to do the next step? Why the result is limited by 2?
| assert not intersect_bfs.accepts(["c", "a"]) | ||
|
|
||
| assert intersect_kron.is_equivalent_to(intersect_bfs) | ||
| assert result.nvals == 2 * len(source_verts) |
There was a problem hiding this comment.
Can we check not only size of result, but also its content?
|
|
||
| return mask_matrix | ||
|
|
||
| def intersect_bfs(self, src_verts) -> EpsilonNFA: | ||
| """ | ||
| Intersection implementation with synchronous breadth first traversal |
There was a problem hiding this comment.
Looks like this function solves reachability problem, not automata intersection. Isn't it?
| # TODO: then here I should assign matrix elements to True | ||
| with Accum(binaryop.MAX_BOOL): | ||
| # extract left (grammar) part of the masks matrix and rearrange rows | ||
| i_x, i_y, _ = found.extract_matrix(col_index=slice(0, num_vert_regex - 1)).to_lists() |
There was a problem hiding this comment.
Can extract_matrix be replaced with Python array slicing?
No description provided.