doc(udf): merge udf docs for the same function

[aceforeverd]

some UDFs may has mutiple doc string, each for different parameters type. The problem is for udf doc gen, it is unclear which doc string to publish in the generated doc (as the document only has one entry for each, it selects random one). Refer the PR #2734 .

know problem udfs

• ilike_match
• like_match
• regexp_like
• string
• substring
• or anything else you find

Solution

merge the multiple doc strings into single one, with all informations included. See how dayofyear implemented:

OpenMLDB/hybridse/src/udf/default_udf_library.cc

Lines 2178 to 2210 in c1d5e1d

	const std::string dayofyear_doc =
	R"(
	@brief Return the day of year for a timestamp or date. Returns 0 given an invalid date.
	Example:
	@code{.sql}
	select dayofyear(timestamp(1590115420000));
	-- output 143
	select dayofyear(1590115420000);
	-- output 143
	select dayofyear(date("2020-05-22"));
	-- output 143
	select dayofyear(date("2020-05-32"));
	-- output 0
	@endcode
	@since 0.1.0
	)";
	RegisterExternal("dayofyear")
	.args<int64_t>(reinterpret_cast<void*>(static_cast<void (*)(int64_t, int32_t*, bool*)>(v1::dayofyear)))
	.return_by_arg(true)
	.returns<Nullable<int32_t>>()
	.doc(dayofyear_doc);
	RegisterExternal("dayofyear")
	.args<Timestamp>(reinterpret_cast<void*>(static_cast<void (*)(Timestamp*, int32_t*, bool*)>(v1::dayofyear)))
	.return_by_arg(true)
	.returns<Nullable<int32_t>>()
	.doc(dayofyear_doc);

[AndyKhang404]

hello, can i work on this issue?

[aceforeverd]

hello, can i work on this issue?

@AndyKhang404 of course you can.

[AndyKhang404]

Thanks for your reply, @aceforeverd. I have some questions:

• What I understand from the issue and after reading the code is that some functions (like regexp_like) can accept multiple types of parameters, but does basically the same thing. My job is to extract the docstrings and merge them together, so that they cover all the cases. Is that correct?
• After merging the docs, are there any tests that I need to run to be sure that I don't accidentally break something else in the process?

This is my first issue, so if there's anything else you'd like to add, please do.

[aceforeverd]

Thanks for your reply, @aceforeverd. I have some questions:

• What I understand from the issue and after reading the code is that some functions (like regexp_like) can accept multiple types of parameters, but does basically the same thing. My job is to extract the docstrings and merge them together, so that they cover all the cases. Is that correct?

yes

• After merging the docs, are there any tests that I need to run to be sure that I don't accidentally break something else in the process?

no, it will not break test

This is my first issue, so if there's anything else you'd like to add, please do.

Since the UDF document introduce functions exactly by function name, all parameter types of same function name are merged into single place. So eventually one document exists for one function, if multiple documents offered in source code, it pick arbitrary one, which is, undetermined.

Sorry for the late.