[
https://issues.apache.org/jira/browse/SOLR-11093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16100480#comment-16100480
]
David Smiley commented on SOLR-11093:
-------------------------------------
Why is zeroCount an integer instead of simply a hasZero boolean?
Here's my first iteration (subsequently changed futher below). All I did was
rename {{hasNext}} to {{positioned}} (which I feel is more clear), added some
inline comments for what the fields mean, and I added a leading check inside
next(). So this was simpler than I thought.
{code:java}
/** Returns an iterator over the values in the set. */
public LongIterator iterator() {
return new LongIterator() {
// if this set contains zero, this iterator's initial state is already
positioned
private boolean positioned = zeroCount > 0;
private int i = -1; // current index into vals[]
private long value = 0; // if positioned, this is our current value
@Override
public boolean hasNext() {
if (positioned) {
return true;
}
while (++i < vals.length) {
value = vals[i];
if (value != 0) {
return positioned = true;
}
}
return false;
}
@Override
public long next() {
if (!positioned) {
if (!hasNext()) {
throw new NoSuchElementException();
}
}
positioned = false;
return value;
}
};
}
{code}
And here's what I think is the most clear approach using a countdown integer:
{code:java}
/** Returns an iterator over the values in the set. */
public LongIterator iterator() {
return new LongIterator() {
private int remainingValues = cardinality();
private int valsIdx = 0;
@Override
public boolean hasNext() {
return remainingValues > 0;
}
@Override
public long next() {
if (!hasNext()) {
throw new NoSuchElementException();
}
remainingValues--;
if (remainingValues == 0 && zeroCount > 0) {
return 0;
}
while (true) { // guaranteed to find another value if we get here
long value = vals[valsIdx++];
if (value != 0) {
return value;
}
}
}
};
{code}
This has the benefit that we don't loop past the last value of the array. It
also has fewer state variables. Also, you don't "pay" any advancement cost in
hasNext(); it's only next() where the value is going to actually be consumed.
> implement Points/Numeric support for graph query
> ------------------------------------------------
>
> Key: SOLR-11093
> URL: https://issues.apache.org/jira/browse/SOLR-11093
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Yonik Seeley
> Assignee: Yonik Seeley
> Fix For: 7.0
>
> Attachments: SOLR-11093.patch, SOLR-11093.patch, SOLR-11093.patch
>
>
> It looks like GraphQueryTest only has tests for strings. We should add tests
> for numeric fields and then enable points randomization.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
