[Optimization] Init map size

[Oliverwqcwrw]

No description provided.

[coveralls]

Coverage remained the same at 87.299% when pulling 1a9e632 on Oliverwqcwrw:master-improve-performance-8 into a68ba4b on mybatis:master.

[harawata]

Hello @Oliverwqcwrw ,

Thank you, but the effect of this change may be minimal.
If you have data that proves otherwise, please post it along with the code that was used to measure the difference.

[Oliverwqcwrw]

My idea is initialized by size to prevent namespaceDiscriminatorMap expansion, because the discriminator case node more than a dozen will trigger the expansion
e.g.

   <discriminator javaType="String" column="did">
        <case value="101" resultType="empInfo">
            <result column="ename" property="name"/>
        </case>
        <case value="102" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
        <case value="103" resultType="empInfo">
            <result column="eage" property="age"/>
        </case>
        <case value="104" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
        <case value="105" resultType="empInfo">
            <result column="ephone" property="phone"/>
        </case>
        <case value="106" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
        <case value="107" resultType="empInfo">
            <result column="ename" property="name"/>
        </case>
        <case value="108" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
        <case value="109" resultType="empInfo">
            <result column="eage" property="age"/>
        </case>
        <case value="110" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
        <case value="111" resultType="empInfo">
            <result column="ephone" property="phone"/>
        </case>
        <case value="112" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
        <case value="113" resultType="empInfo">
            <result column="ename" property="name"/>
        </case>
        <case value="114" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
        <case value="115" resultType="empInfo">
            <result column="eage" property="age"/>
        </case>
        <case value="116" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
        <case value="117" resultType="empInfo">
            <result column="ephone" property="phone"/>
        </case>
        <case value="118" resultType="empInfo">
            <result column="ejob" property="job"/>
        </case>
    </discriminator>

[harawata]

I understand your intent, but building mapper occurs only once during application startup and saving a few extra expansion does not make much difference in reality, I think.
As I wrote, if you have data that shows the effect of this change, I would be happy to reconsider.

[Oliverwqcwrw]

I have tested the situation of expansion and no expansion, and the data is as follows

Expansion takes nanoTime = 83521
Without expansion takes nanoTime = 8184

Just as the data, the difference is quite big, although the impact of one expansion is relatively small, but in reality, we may have hundreds or even more mapper for business development, if each one saves some expansion time, then the growth of our startup speed will be obvious

[harawata]

Could you share the project you used to measure the time?

[Oliverwqcwrw]

@Test
public void test(){
    HashMap<String, String> map = new HashMap<>();
    final long start1 = System.nanoTime();
    for (int i = 0; i < 20; i++) {
        map.put("key" + i, "value");
    }
    System.out.println("Expansion takes nanoTime = " + (System.nanoTime() - start1));

    HashMap<String,String> map2 = new HashMap<>(20, 1);
    final long start2 = System.nanoTime();
    for (int i = 0; i < 20; i++) {
        map2.put("key" + i, "value");
    }
    System.out.println("Without expansion takes nanoTime = " + (System.nanoTime() - start2));
}

[harawata]

That really is not the right way to measure performance and the result should be rather random.

So, you are not actually experiencing the performance issue with your project, is that correct?
I appreciate you taking the time to submit a PR, but this type of micro-optimization should be done only when there is an actual problem.
There are some common advices on this question:
https://softwareengineering.stackexchange.com/q/99445

In this particular case, if there are so many discriminators in a result map that map expansion is triggered repeatedly, there may be some kind of design problem and you'll experience real performance issue when mapping the result, I think.

[Oliverwqcwrw]

Thanks for your reply,I agree with you