fix: Outdated image string join logic in ImageRow.image_ref

[jopemachine]

Fixes #3126, follow-up to #2978.

Since project value was changed to be nullable in PR #2978, the image string join logic also should be updated.

Why?

For example, if the registry is "abc" and the project is "", then f"{self.registry}/{self.project}/" becomes "abc//". (Expected result: "abc/")

---

Checklist: (if applicable)

• Milestone metadata specifying the target backport version

[fregataa]

Good. After this PR, let's refactor this codes by replacing split() to partition() and adding type hints to the join_non_empty() function

[jopemachine]

Good. After this PR, let's refactor this codes by replacing split() to partition() and adding type hints to the join_non_empty() function

@fregataa You might be referring to the following code.
By the way, is there a specific reason we prefer partition() over split()?
I don't see significant readability difference between code that uses split() and code that uses partition().

image_and_tag = self.name.partition(f"{join(self.registry, self.project)}/")[2]

[fregataa]

@fregataa You might be referring to the following code. By the way, is there a specific reason we prefer partition() over split()? I don't see significant readability difference between code that uses split() and code that uses partition().

image_and_tag = self.name.partition(f"{join(self.registry, self.project)}/")[2]

partition() always returns a tuple of three strings but split() returns a list of strings.
you can use split() if you expect the string value will be divided into more than 2 parts. but if you are sure that the value is divided into 2 parts, using partition() makes it easy to understand how the result comes from the original value.

image_and_tag = self.name.partition(f"{join(self.registry, self.project)}/")[2] # This is okay but
_, _, image_and_tag = self.name.partition(f"{join(self.registry, self.project)}/") # I recommend to use this pattern with partition()

[jopemachine]

@fregataa You might be referring to the following code. By the way, is there a specific reason we prefer partition() over split()? I don't see significant readability difference between code that uses split() and code that uses partition().

image_and_tag = self.name.partition(f"{join(self.registry, self.project)}/")[2]

partition() always returns a tuple of three strings but split() returns a list of strings. you can use split() if you expect the string value will be divided into more than 2 parts. but if you are sure that the value is divided into 2 parts, using partition() makes it easy to understand how the result comes from the original value.

image_and_tag = self.name.partition(f"{join(self.registry, self.project)}/")[2] # This is okay but
_, _, image_and_tag = self.name.partition(f"{join(self.registry, self.project)}/") # I recommend to use this pattern with partition()

Thanks for your description! I will reflect this.