hadoop

At this moment relu_layer op doesn't allow threshold configuration, and legacy RELU op allows that.
We should add configuration option to relu_layer.

We should use MockConnector to simulate connector that support missing columns on insert and test if row filtering is properly handled if columns are missing during the insert but used in row filters.

Related to: #10006 and #10005

问题描述：关于工程中某个工作流节点单独执行不报错，保存工作流时提示：NotClassDefFoundError：Cloud not initialize class dispatch.Http$

问题原因：linkis-publish微服务关于netty-3.6.2.Final.jar升级包缺失

问题解决：将升级包上传重启linkis-publish微服务即可。

Append class to all HashCodeBuilders in Gaffer for the below issue to minimise hash collisions.

@Test
    void name() {
        Foo foo = new Foo();
        Bar bar = new Bar();

        assertFalse(foo.equals(bar));
        assertNotEquals(foo.hashCode(), bar.hashCode()); //fails
    }

    class Bar {
        int a = 3;

        @Override
        public int hashCode() {

When constructing expressions involving a non-nullable type and a NULL value, ibis is unable to compute precedence for the type:

In [10]: import ibis

In [11]: import ibis.expr.datatypes as dt

In [12]: t = ibis.table([("a", dt.Int64(nullable=False))], name="t")

In [13]: t.a.coalesce(ibis.NA)
IbisTypeError: Cannot compute precedence for int64[non-nullable] and null types