Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix testing tools NaN comparison and correct new failing tests. #4794

Merged
merged 11 commits into from
Nov 16, 2023
Merged

Fix testing tools NaN comparison and correct new failing tests. #4794

Show file tree
Hide file tree
Changes from 9 commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
16c7288
Initial commit of emstd bug fix and test.
lbooker42 Nov 7, 2023
f871ff8
Initial commit of NaN test tooling bug-fix.
lbooker42 Nov 8, 2023
cff18be
Merged with main
lbooker42 Nov 8, 2023
916946b
Correct post-merge test failure.
lbooker42 Nov 8, 2023
53a0a64
Some PR comments addressed
lbooker42 Nov 9, 2023
43ddf91
Numeric class modified to better handle NaN and Inf values, new tests…
lbooker42 Nov 10, 2023
d547db3
Expanded the NaN and Inf test code to document our expectations.
lbooker42 Nov 10, 2023
87a50f0
Addressed PR comments.
lbooker42 Nov 16, 2023
47d30ff
Added new tests to cover some short circuit behavior.
lbooker42 Nov 16, 2023
e2b1f18
Update engine/function/src/templates/Numeric.ftl
lbooker42 Nov 16, 2023
7bb5e68
Update engine/function/src/templates/Numeric.ftl
lbooker42 Nov 16, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
169 changes: 86 additions & 83 deletions engine/function/src/templates/Numeric.ftl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -365,6 +365,9 @@ public class Numeric {
try ( final ${pt.vectorIterator} vi = values.iterator() ) {
while ( vi.hasNext() ) {
final ${pt.primitive} c = vi.${pt.iteratorNext}();
if (isNaN(c)) {
return Double.NaN;
}
if (!isNull(c)) {
sum += c;
count++;
Expand Down Expand Up @@ -416,6 +419,12 @@ public class Numeric {
try ( final ${pt.vectorIterator} vi = values.iterator() ) {
while ( vi.hasNext() ) {
final ${pt.primitive} c = vi.${pt.iteratorNext}();
if (isNaN(c)) {
lbooker42 marked this conversation as resolved.
Show resolved Hide resolved
return Double.NaN;
}
if (isInf(c)) {
return Double.POSITIVE_INFINITY;
}
if (!isNull(c)) {
sum += Math.abs(c);
count++;
Expand Down Expand Up @@ -472,11 +481,13 @@ public class Numeric {

double sum = 0;
double sum2 = 0;
double count = 0;

int count = 0;
lbooker42 marked this conversation as resolved.
Show resolved Hide resolved
try ( final ${pt.vectorIterator} vi = values.iterator() ) {
while ( vi.hasNext() ) {
final ${pt.primitive} c = vi.${pt.iteratorNext}();
if (isNaN(c) || isInf(c)) {
return Double.NaN;
}
if (!isNull(c)) {
sum += (double)c;
sum2 += (double)c * (double)c;
Expand All @@ -485,19 +496,19 @@ public class Numeric {
}
}

// Return NaN if poisoned or too few values to compute sample variance.
if (count <= 1 || Double.isNaN(sum) || Double.isNaN(sum2)) {
// Return NaN if overflow or too few values to compute variance.
if (count <= 1 || Double.isInfinite(sum) || Double.isInfinite(sum2)) {
return Double.NaN;
}

// Perform the calculation in a way that minimizes the impact of floating point error.
final double eps = Math.ulp(sum2);
final double vs2bar = sum * (sum / count);
final double vs2bar = sum * (sum / (double)count);
final double delta = sum2 - vs2bar;
final double rel_eps = delta / eps;

// Return zero when the sample variance is leq the floating point error.
return Math.abs(rel_eps) > 1.0 ? delta / (count - 1) : 0.0;
return Math.abs(rel_eps) > 1.0 ? delta / ((double)count - 1) : 0.0;
}

<#list primitiveTypes as pt2>
Expand Down Expand Up @@ -590,7 +601,12 @@ public class Numeric {
while (vi.hasNext()) {
final ${pt.primitive} c = vi.${pt.iteratorNext}();
final ${pt2.primitive} w = wi.${pt2.iteratorNext}();

if (isNaN(c) || isInf(c)) {
return Double.NaN;
}
if (isNaN(w) || isInf(w)) {
return Double.NaN;
}
if (!isNull(c) && !isNull(w)) {
sum += w * c;
sum2 += w * c * c;
Expand All @@ -600,8 +616,8 @@ public class Numeric {
}
}

// Return NaN if poisoned or too few values to compute sample variance.
if (count <= 1 || Double.isNaN(sum) || Double.isNaN(sum2) || Double.isNaN(count) || Double.isNaN(count2)) {
// Return NaN if overflow or too few values to compute variance.
if (count <= 1 || Double.isInfinite(sum) || Double.isInfinite(sum2)) {
return Double.NaN;
}

Expand Down Expand Up @@ -1333,6 +1349,12 @@ public class Numeric {
while (v0i.hasNext()) {
final ${pt.primitive} v0 = v0i.${pt.iteratorNext}();
final ${pt2.primitive} v1 = v1i.${pt2.iteratorNext}();
if (isNaN(v0) || isInf(v0)) {
return Double.NaN;
}
if (isNaN(v1) || isInf(v1)) {
return Double.NaN;
}

if (!isNull(v0) && !isNull(v1)) {
sum0 += v0;
Expand Down Expand Up @@ -1421,6 +1443,12 @@ public class Numeric {
while (v0i.hasNext()) {
final ${pt.primitive} v0 = v0i.${pt.iteratorNext}();
final ${pt2.primitive} v1 = v1i.${pt2.iteratorNext}();
if (isNaN(v0) || isInf(v0)) {
return Double.NaN;
}
if (isNaN(v1) || isInf(v1)) {
return Double.NaN;
}

if (!isNull(v0) && !isNull(v1)) {
sum0 += v0;
Expand Down Expand Up @@ -1460,6 +1488,11 @@ public class Numeric {
try ( final ${pt.vectorIterator} vi = values.iterator() ) {
while ( vi.hasNext() ) {
final ${pt.primitive} c = vi.${pt.iteratorNext}();
<#if pt.valueType.isFloat >
if (isNaN(c)) {
return ${pt.boxed}.NaN;
}
</#if>
if (!isNull(c)) {
sum += c;
}
Expand Down Expand Up @@ -1496,10 +1529,33 @@ public class Numeric {

${pt.primitive} prod = 1;
int count = 0;
<#if pt.valueType.isFloat >
int zeroCount = 0;
int infCount = 0;
lbooker42 marked this conversation as resolved.
Show resolved Hide resolved
</#if>

try ( final ${pt.vectorIterator} vi = values.iterator() ) {
while ( vi.hasNext() ) {
final ${pt.primitive} c = vi.${pt.iteratorNext}();
<#if pt.valueType.isFloat >
if (isNaN(c)) {
return ${pt.boxed}.NaN;
} else if (Double.isInfinite(c)) {
if (zeroCount > 0) {
return ${pt.boxed}.NaN;
}
infCount++;
} else if (c == 0) {
if (infCount > 0) {
return ${pt.boxed}.NaN;
}
zeroCount++;
}
<#else>
if (c == 0) {
return 0;
}
</#if>
if (!isNull(c)) {
count++;
prod *= c;
Expand All @@ -1511,7 +1567,11 @@ public class Numeric {
return ${pt.null};
}

<#if pt.valueType.isFloat >
return zeroCount > 0 ? 0 : (${pt.primitive}) (prod);
<#else>
return (${pt.primitive}) (prod);
</#if>
}

/**
Expand Down Expand Up @@ -1549,24 +1609,7 @@ public class Numeric {
return null;
}

if (values.length == 0) {
return new ${pt.primitive}[0];
}

${pt.primitive}[] result = new ${pt.primitive}[values.length];
result[0] = values[0];

for (int i = 1; i < values.length; i++) {
if (isNull(result[i - 1])) {
result[i] = values[i];
} else if (isNull(values[i])) {
result[i] = result[i - 1];
} else {
result[i] = (${pt.primitive})Math.min(result[i - 1], values[i]);
}
}

return result;
return cummin(new ${pt.vectorDirect}(values));
}

/**
Expand Down Expand Up @@ -1630,24 +1673,7 @@ public class Numeric {
return null;
}

if (values.length == 0) {
return new ${pt.primitive}[0];
}

${pt.primitive}[] result = new ${pt.primitive}[values.length];
result[0] = values[0];

for (int i = 1; i < values.length; i++) {
if (isNull(result[i - 1])) {
result[i] = values[i];
} else if (isNull(values[i])) {
result[i] = result[i - 1];
} else {
result[i] = (${pt.primitive})Math.max(result[i - 1], values[i]);
}
}

return result;
return cummax(new ${pt.vectorDirect}(values));
}

/**
Expand Down Expand Up @@ -1711,24 +1737,7 @@ public class Numeric {
return null;
}

if (values.length == 0) {
return new ${pt.primitive}[0];
}

${pt.primitive}[] result = new ${pt.primitive}[values.length];
result[0] = values[0];

for (int i = 1; i < values.length; i++) {
if (isNull(result[i - 1])) {
result[i] = values[i];
} else if (isNull(values[i])) {
result[i] = result[i - 1];
} else {
result[i] = (${pt.primitive}) (result[i - 1] + values[i]);
}
}

return result;
return cumsum(new ${pt.vectorDirect}(values));
}

/**
Expand Down Expand Up @@ -1792,24 +1801,7 @@ public class Numeric {
return null;
}

if (values.length == 0) {
return new ${pt.primitive}[0];
}

${pt.primitive}[] result = new ${pt.primitive}[values.length];
result[0] = values[0];

for (int i = 1; i < values.length; i++) {
if (isNull(result[i - 1])) {
result[i] = values[i];
} else if (isNull(values[i])) {
result[i] = result[i - 1];
} else {
result[i] = (${pt.primitive}) (result[i - 1] * values[i]);
}
}

return result;
return cumprod(new ${pt.vectorDirect}(values));
}

/**
Expand Down Expand Up @@ -2322,7 +2314,13 @@ public class Numeric {
while (vi.hasNext()) {
final ${pt.primitive} c = vi.${pt.iteratorNext}();
final ${pt2.primitive} w = wi.${pt2.iteratorNext}();

if (isNaN(c)) {
return Double.NaN;
}
if (isNaN(w)) {
return Double.NaN;
}

if (!isNull(c) && !isNull(w)) {
vsum += c * w;
}
Expand Down Expand Up @@ -2405,7 +2403,12 @@ public class Numeric {
while (vi.hasNext()) {
final ${pt.primitive} c = vi.${pt.iteratorNext}();
final ${pt2.primitive} w = wi.${pt2.iteratorNext}();

if (isNaN(c)) {
return Double.NaN;
}
if (isNaN(w)) {
return Double.NaN;
}
if (!isNull(c) && !isNull(w)) {
vsum += c * w;
wsum += w;
Expand Down
Loading
Loading