## Highlight Exact Source Phrases and Attach Citations You are given a text string and an indexed list of source phrases. All input tokens are alphanumeric and separated by single spaces. A source matches only as a sequence of whole words; for example, `network` does not match `networking`. Implement both stages below. ### Stage 1 - Merged Highlights Find every exact occurrence of every source phrase. Convert each occurrence to a half-open word interval. Merge intervals that overlap; do not merge intervals separated by one or more unmatched words. Wrap each merged interval in `<yellow>` and `</yellow>` tags. Example: ```text text = "the quick brown fox" sources = ["the quick", "quick brown"] result = "<yellow>the quick brown</yellow> fox" ``` The two source occurrences overlap at `quick`, so they form one properly nested highlight rather than crossing tags. ### Stage 2 - Citation Ordering After each merged highlight, append one citation `[i]` for every distinct source index with at least one occurrence overlapping that highlight. Order those citation indices by: 1. The source's total number of occurrences across the entire input text, descending. 2. Source index ascending when total counts tie. Repeated occurrences of one source inside the same merged region still produce that source's citation only once for that region. Example: ```text text = "brown v board changed law brown v board" sources = ["brown v board changed", "brown v board", "changed law"] ``` Source 1 occurs twice; sources 0 and 2 occur once. The first merged region therefore ends with citations in the order `[1][0][2]`. The second region has only `[1]`. ### Function Contract ```text highlight_with_citations(text, sources) -> string ``` Return the Stage 2 result, including tags and citations. If there are no matches, return `text` unchanged. ### Constraints - `0 <= number of words in text <= 10000` - `0 <= len(sources) <= 1000` - Every source contains at least one word. - A source may occur multiple times, including overlapping with itself. - Duplicate source strings at different indices remain distinct citation sources. - Matching is case-sensitive. ### Hints - Tokenize once and record occurrences as `(start, end, source_index)`. - Sort intervals by start position, then merge while retaining the set of contributing source indices. - Compute global occurrence counts before rendering citations. - Render from the word sequence so tags never split a word. ### Discussion Extensions - Analyze the straightforward matching complexity and suggest an approach for many source phrases. - How would punctuation and original whitespace preservation change tokenization and rendering? - What behavior would you choose for adjacent, non-overlapping highlights?