Hi, I'm considering writing a plugin to generate a...

# dokka

[solonovamax]

Hi,
I'm considering writing a plugin to generate a package overview for my kotlin module.
Currently, I have the following:

| Name                                            | Distance | Similarity | Normalized | Metric | Memory cost          | Execution cost                      | Typical usage   |
|-------------------------------------------------|:--------:|:----------:|:----------:|:------:|----------------------|-------------------------------------|-----------------|
| Levenshtein                                     |    ☒     |     ☐      |     ☐      |   ☒    | \\(O(m \\times n)\\) | \\(O(m \\times n)\\) <sup>1</sup>   |                 |
| Damerau-Levenshtein<sup>3</sup>                 |    ☒     |     ☐      |     ☐      |   ☒    | \\(O(m \\times n)\\) | \\(O(m \\times n)\\) <sup>1</sup>   |                 |
| Optimal String Alignment<sup>3</sup>            |    ☒     |     ☐      |     ☐      |   ☒    | \\(O(m \\times n)\\) | \\(O(m \\times n)\\) <sup>1</sup>   |                 |
| Longest Common Subsequence                      |    ☒     |     ☐      |     ☐      |   ☒    | \\(O(m \\times n)\\) | \\(O(m \\times n)\\) <sup>1,2</sup> | diff, git       |
| Normalized Levenshtein                          |    ☒     |     ☒      |     ☒      |   ☒    | \\(O(m \\times n)\\) | \\(O(m \\times n)\\) <sup>1</sup>   |                 |
| Normalized Damerau-Levenshtein<sup>3</sup>      |    ☒     |     ☐      |     ☒      |   ☒    | \\(O(m \\times n)\\) | \\(O(m \\times n)\\) <sup>1</sup>   |                 |
| Normalized Optimal String Alignment<sup>3</sup> |    ☒     |     ☐      |     ☒      |   ☒    | \\(O(m \\times n)\\) | \\(O(m \\times n)\\) <sup>1</sup>   |                 |
| Normalized Longest Common Subsequence           |    ☒     |     ☐      |     ☒      |   ☒    | \\(O(m \\times n)\\) | \\(O(m \\times n)\\) <sup>1,2</sup> |                 |
| Cosine similarity                               |    ☒     |     ☒      |     ☒      |   ☐    | \\(O(m + n)\\)       | \\(O(m + n)\\)                      |                 |
| Jaccard index                                   |    ☒     |     ☒      |     ☒      |   ☒    | \\(O(m + n)\\)       | \\(O(m + n)\\)                      |                 |
| Jaro-Winkler                                    |    ☒     |     ☒      |     ☒      |   ☐    | \\(O(m + n)\\)       | \\(O(m \\times n)\\)                | typo correction |
| N-Gram                                          |    ☒     |     ☐      |     ☒      |   ☐    |                      | \\(O(m \\times n)\\)                |                 |
| Q-Gram                                          |    ☒     |     ☐      |     ☐      |   ☐    |                      | \\(O(m + n)\\)                      |                 |
| Ratcliff-Obershelp                              |    ☒     |     ☒      |     ☒      |   ☐    | \\(O(m + n)\\)       | \\(O(n^3)\\)                        |                 |
| Sorensen-Dice coefficient                       |    ☒     |     ☒      |     ☒      |   ☐    |                      | \\(O(m + n)\\)                      |                 |
| Sift 4                                          |    ☒     |     ☐      |     ☐      |   ☐    | \\(O(m + n)\\)       | \\(O(m + n)\\)                      |                 |

this has to be updated by-hand whenever a new algorithm is added.
I was wondering if there would be a good way that I could have a plugin that goes through all the classes that implement a specific interface/interfaces (in this case StringDistance and StringSimilarity and record them, then aggregate all the information in the index html file.

If yes, what would be the best places to hook into for this? (I'm assuming that things like the memory cost/execution cost can be added with a custom tag, eg. @memoryComplexity and @timeComplexity, for example:

/**
 * Sift4 distance, as defined by
 * Manda, Costin, [Siderite]. "Super Fast and Accurate String Distance Algorithm: Sift4." Siderite’s Blog,
 * 10 Nov. 2014, <https://siderite.dev/blog/super-fast-and-accurate-string-distance.html>.
 *
 * Note: this algorithm is asymmetric. This means that
 * \(distance(X, Y) \not\equiv distance(Y, X)\).
 * This is one of the artifacts of the linear nature of the algorithm.
 *
 * @memoryComplexity O(m + n)
 * @timeComplexity O(m + n)
 * @author Thibault Debatty, solonovamax
 */
@ExperimentalStringMeasure
public class Sift4(
    /**
     * Set the maximum distance to search for character transposition.
     * Compute cost of algorithm is `O(n * maxOffset)`
     */
    private val maxOffset: Int = DEFAULT_MAX_OFFSET,
) : StringDistance {

[glureau]

I'd go with a KSP plugin to do that:
• provides some annotations for your measurements
• KSP produces a markdown file for a given package/module (this file will go in /build/)
• You may then use a tool like grip (disclaimer: I'm the author) to replace a directive from your dokka module markdown file with the output of KSP
With this kind of approach, the generated markdowns can also be used with other tools than dokka (like github pages, a simple mkdocs website, etc.). Also it demands a bit more setup to plug the files into dokka, so that's eventually your choice, just wanted to show another approach 🙂