I am a software developer from China. Recently I am reading the source of lucene of 6.6.0. I can not understand some code of FST. List them below.
1 . There is a field called 'inCounts' in the class of FST. But in my opinion every node of FST has only one arc pointing to it. If a node is not shared, it is certain it has only one arc pointing to it since it is like a tree. If a node is shared,it is returned directly from NodeHash so the number of it's coming in arc is not updated. So I just can't understand this attribute though it is used is 'pach' method.
2. Is about the method of 'pach' of class FST. This line 'long delta = newNodeAddress.get((int) arc.target) + addressError - writer.getPosition() - 2;' . I know it is comparing which one is smaller but I just can't understand why 2 is minused.
3. If a FST is packed,then there are 3 ways to save the target of an arc, first is the difference of the target and this arc ,second is the ordinal number from a hashmap if the target is among those having top N incoming arcs, third is absolute position of the node. But in the method 'readNextRealArc' , I can not understand the process of the absolute position while reading. Below is the code.
if (packed) {
final long pos = in.getPosition();
final long code = in.readVLong();
if (arc.flag(BIT_TARGET_DELTA)) {
// record the difference
} else if (code < nodeRefToAddress.size()) {
// record the ordinal number from a hashmap
} else {// Absolute position
arc.target = code;//I can not understa this.In opinion, nodeRefToAddress.size() should be minused because it is summed into the absolute value in 'pack' method.
}
} else {
-----------------------
Please pardon me if I use some wrong words ,my English is not good.Thanks for reading the email. Looking forward to the reply.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
