Linas Vepštas
Linas Vepštas
Yes, send a pull req. At each location, add comments such as `/* can add at most ((2^31-1) * (2^31-1)) < 2^62, so the result is still less than 2^63...
> billions of disjuncts per word, That won't happen. There won't even be a million. However, in the earlier days, I had SQL dicts that had a hundred `UNKNOWN-WORD`, each...
Ah interesting. It certainly is very slow, on that sentence, and then shoots up to 6.6GB on my machine. So, yes, this is an example where the current dictionaries are...
> a loop over a large number of disjuncts/connectors Where is that loop? It would be better if that loop was a part of the dictionary setup... > it may...
* OK longjmp is easier than exceptions. * I'm confused. I think you are saying that you want to add an LL link-length limit to `set_connector_length_limits`, right? Because there isn't...
I have no objection to these changes, they seem reasonable.
2 or 3 or 4 are reasonable. Given that this bug was opened less than a year ago, option 2 or 4 seems like the the "nice" thing to do....
Yes, the `YS` and `YP` and `PH` connectors in English are all length=1
I just remove xalloc from the post-processing code
`valgrind --leak-check=full link-parser < ../data/en/corpus-basic.batch` says ``` ==25945== HEAP SUMMARY: ==25945== in use at exit: 0 bytes in 0 blocks ==25945== total heap usage: 54,015,957 allocs, 54,015,957 frees, 2,313,157,426 bytes...