Text search should support context

[jrieken]

We should consider something like the --context option (https://github.com/BurntSushi/ripgrep/blob/master/GUIDE.md#common-options) instead of the current preview options. They currently don't allow to show a multi line context around a match and make implementing a textual representation of search results very hard...

[roblourens]

I would add

export interface TextSearchPreviewOptions {
    contextLines?: number;
}

[jrieken]

Yeah, that would be nice but might also be problematic. When having context: boolean the search engine can decide, e.g. show 5 lines (two above & below) for normal matches but only show parts of a line for monster lines

[roblourens]

Sometimes it's up to the user how much context should be shown. In Atom I can change leading and trailing context via the UI independently.

If you are showing results as text with context, is it annoying that multiple results on one line are given as multiple result objects, possibly with overlapping previews? Should the API keep the one-result-per-line model instead?

One difficulty here is when you have a long line with one match at the beginning and one at the end.

But to extend the current model to that I think would only require us to change the two ranges to Range|Range[]

[jrieken]

Should the API keep the one-result-per-line model instead?

That means to show more context (line above/below) the file needs to opened and I don't that feasible...

[roblourens]

I'm referring to something else, not context lines. When you have two matches on the same line, the API will send two result objects. For your extension, it might be better if you can get a single result object that indicates the two matches. That's what I mean by one result per line.

Kai had the suggestion of just using exactly ripgrep's format: https://burntsushi.net/stuff/libripgrep/doc/grep_printer/struct.JSON.html

This is all kind of tangential but related because it probably doesn't make sense to show the same context lines twice when there are two matches on one line.

[roblourens]

Opened #59919 for the other topic, the API change for context would look like this:

export interface TextSearchOptions extends SearchOptions {
        // ...

	/**
	 * Lines of context after the match
	 */
	afterContext?: number;

	/**
	 * Lines of context before the match
	 */
	beforeContext?: number;
}

export interface TextSearchContext {
        /**
	 * The uri for the matching document.
         */
	uri: Uri;

	/**
	 * One or more lines of text.
	 * previewOptions.charsPerLine applies to this
	 */
	text: string;

	/**
	 * The range of this context within the document.
	 */
	range: Range;
}

export type TextSearchResult = TextSearchLineMatch | TextSearchContext;

I think we would sort all of these results on the frontend.

[roblourens]

Please verify by calling the findTextInFiles API from an extension with the context options, check that you get something reasonable.