Draft: proof-of-concept for Java 17 API improvements #19
Conversation
New classes are in the net.ripe.ipresource.scratch package. Major changes are: 1. Single resources (ASN, IPv4 address, IPv6 address) no longer extends the range classes. 2. Additional implementation classes for IP prefixes (both v4 and v6) so you can use static types when a prefix is required. 3. IPv6 address now uses two `long`s to represent the address, instead of a `BigInteger`. 4. ASN ranges and IPv4 blocks use primitive types internally to reduce memory usage. This does result in allocations when converting the start and end points into ASN/IPv4Address instances. 5. Internal code tries to avoid allocations in common paths, prefering to use primitive types instead. 6. Old RIPE prefix notation without trailing zeroes (e.g. 10.0/16) is not supported. Still many things to do: 1. Documentation and tests. 2. Code makes use of the preview pattern matching feature, for production release this should be replaced with if-else chains. 3. Parsing code needs to be ported from old classes. 4. API is not complete yet.
|
Also, there are no separate classes for |
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
There was a problem hiding this comment.
CodeQL found more than 10 potential problems in the proposed changes. Check the Files changed tab for more details.
|
Should we rename |
| private static Ipv4Block of(long start, long end) { | ||
| if (isLegalPrefix(start, end)) { | ||
| long temp = start ^ end; | ||
| int length = Integer.numberOfLeadingZeros((int) temp); |
There was a problem hiding this comment.
Do I understand correctly that it's irreversible? So if the (being, end) range is not a prefix, it will build the biggest prefix fitting into the range?
| import static net.ripe.ipresource.scratch.Ipv6Address.parse; | ||
| import static org.junit.Assert.*; | ||
|
|
||
| public class Ipv6AddressTest { |
There was a problem hiding this comment.
Can we have a property test for parse(toString(i)) == i?
|
It would be nice to fix the CodeQL-found issues |
Upgrade JUnit to allow parameterized testing.
|
|
||
| public final class Ipv6Address implements IpAddress { | ||
| public static final int NUMBER_OF_BITS = 128; | ||
|
|
||
| public static final Ipv6Address LOWEST = new Ipv6Address(0, 0); | ||
| public static final Ipv6Address HIGHEST = new Ipv6Address(-1, -1); | ||
|
|
||
| /* Pattern to match IPv6 addresses in forms defined in http://www.ietf.org/rfc/rfc4291.txt */ |
There was a problem hiding this comment.
It looks quite intimidating and complex.
Don't we want to use some library for parsing?
https://github.com/AdoptOpenJDK/openjdk-jdk11/blob/master/src/java.base/share/classes/sun/net/util/IPAddressUtil.java
Or Guava or something
New classes are in the net.ripe.ipresource.scratch package.
Major changes are:
Single resources (ASN, IPv4 address, IPv6 address) no longer extends the range classes.
Additional implementation classes for IP prefixes (both v4 and v6) so you can use static types when a prefix is required.
IPv6 address now uses two
longs to represent the address, instead of aBigInteger.ASN ranges and IPv4 blocks use primitive types internally to reduce memory usage. This does result in allocations when converting the start and end points into ASN/IPv4Address instances.
Internal code tries to avoid allocations in common paths, prefering to use primitive types instead.
Old RIPE prefix notation without trailing zeroes (e.g. 10.0/16) is not supported.
Still many things to do:
Documentation and tests.
Code makes use of the preview pattern matching feature, for production release this should be replaced with if-else chains.
Parsing code needs to be ported from old classes.
API is not complete yet.