|
|
|
@ -52,9 +52,65 @@ const removeMarkdown = (
|
|
|
|
|
return markdown
|
|
|
|
|
}
|
|
|
|
|
return output
|
|
|
|
|
};
|
|
|
|
|
}
|
|
|
|
|
// -----
|
|
|
|
|
|
|
|
|
|
const highlight = (content, term) => {
|
|
|
|
|
const highlightWindow = 20
|
|
|
|
|
|
|
|
|
|
// try to find direct match first
|
|
|
|
|
const directMatchIdx = content.indexOf(term)
|
|
|
|
|
if (directMatchIdx !== -1) {
|
|
|
|
|
const h = highlightWindow / 2
|
|
|
|
|
const before = content.substring(0, directMatchIdx).split(" ").slice(-h)
|
|
|
|
|
const after = content.substring(directMatchIdx + term.length, content.length - 1).split(" ").slice(0, h)
|
|
|
|
|
return (before.length == h ? `...${before.join(" ")}` : before.join(" ")) + `<span class="search-highlight">${term}</span>` + after.join(" ")
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const tokenizedTerm = term.split(/\s+/).filter((t) => t !== '')
|
|
|
|
|
const splitText = content.split(/\s+/).filter((t) => t !== '')
|
|
|
|
|
const includesCheck = (token) =>
|
|
|
|
|
tokenizedTerm.some((term) =>
|
|
|
|
|
token.toLowerCase().startsWith(term.toLowerCase())
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
const occurrencesIndices = splitText.map(includesCheck)
|
|
|
|
|
|
|
|
|
|
// calculate best index
|
|
|
|
|
let bestSum = 0
|
|
|
|
|
let bestIndex = 0
|
|
|
|
|
for (
|
|
|
|
|
let i = 0;
|
|
|
|
|
i < Math.max(occurrencesIndices.length - highlightWindow, 0);
|
|
|
|
|
i++
|
|
|
|
|
) {
|
|
|
|
|
const window = occurrencesIndices.slice(i, i + highlightWindow)
|
|
|
|
|
const windowSum = window.reduce((total, cur) => total + cur, 0)
|
|
|
|
|
if (windowSum >= bestSum) {
|
|
|
|
|
bestSum = windowSum
|
|
|
|
|
bestIndex = i
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const startIndex = Math.max(bestIndex - highlightWindow, 0)
|
|
|
|
|
const endIndex = Math.min(
|
|
|
|
|
startIndex + 2 * highlightWindow,
|
|
|
|
|
splitText.length
|
|
|
|
|
)
|
|
|
|
|
const mappedText = splitText
|
|
|
|
|
.slice(startIndex, endIndex)
|
|
|
|
|
.map((token) => {
|
|
|
|
|
if (includesCheck(token)) {
|
|
|
|
|
return `<span class="search-highlight">${token}</span>`
|
|
|
|
|
}
|
|
|
|
|
return token
|
|
|
|
|
})
|
|
|
|
|
.join(' ')
|
|
|
|
|
.replaceAll('</span> <span class="search-highlight">', ' ')
|
|
|
|
|
return `${startIndex === 0 ? '' : '...'}${mappedText}${endIndex === splitText.length ? '' : '...'
|
|
|
|
|
}`
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
(async function() {
|
|
|
|
|
const encoder = (str) => str.toLowerCase().split(/([^a-z]|[^\x00-\x7F])+/)
|
|
|
|
|
const contentIndex = new FlexSearch.Document({
|
|
|
|
@ -84,52 +140,6 @@ const removeMarkdown = (
|
|
|
|
|
})
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const highlight = (content, term) => {
|
|
|
|
|
const highlightWindow = 20
|
|
|
|
|
const tokenizedTerm = term.split(/\s+/).filter((t) => t !== '')
|
|
|
|
|
const splitText = content.split(/\s+/).filter((t) => t !== '')
|
|
|
|
|
const includesCheck = (token) =>
|
|
|
|
|
tokenizedTerm.some((term) =>
|
|
|
|
|
token.toLowerCase().startsWith(term.toLowerCase())
|
|
|
|
|
)
|
|
|
|
|
|
|
|
|
|
const occurrencesIndices = splitText.map(includesCheck)
|
|
|
|
|
|
|
|
|
|
// calculate best index
|
|
|
|
|
let bestSum = 0
|
|
|
|
|
let bestIndex = 0
|
|
|
|
|
for (
|
|
|
|
|
let i = 0;
|
|
|
|
|
i < Math.max(occurrencesIndices.length - highlightWindow, 0);
|
|
|
|
|
i++
|
|
|
|
|
) {
|
|
|
|
|
const window = occurrencesIndices.slice(i, i + highlightWindow)
|
|
|
|
|
const windowSum = window.reduce((total, cur) => total + cur, 0)
|
|
|
|
|
if (windowSum >= bestSum) {
|
|
|
|
|
bestSum = windowSum
|
|
|
|
|
bestIndex = i
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const startIndex = Math.max(bestIndex - highlightWindow, 0)
|
|
|
|
|
const endIndex = Math.min(
|
|
|
|
|
startIndex + 2 * highlightWindow,
|
|
|
|
|
splitText.length
|
|
|
|
|
)
|
|
|
|
|
const mappedText = splitText
|
|
|
|
|
.slice(startIndex, endIndex)
|
|
|
|
|
.map((token) => {
|
|
|
|
|
if (includesCheck(token)) {
|
|
|
|
|
return `<span class="search-highlight">${token}</span>`
|
|
|
|
|
}
|
|
|
|
|
return token
|
|
|
|
|
})
|
|
|
|
|
.join(' ')
|
|
|
|
|
.replaceAll('</span> <span class="search-highlight">', ' ')
|
|
|
|
|
return `${startIndex === 0 ? '' : '...'}${mappedText}${endIndex === splitText.length ? '' : '...'
|
|
|
|
|
}`
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
const resultToHTML = ({ url, title, content, term }) => {
|
|
|
|
|
const text = removeMarkdown(content)
|
|
|
|
|
const resultTitle = highlight(title, term)
|
|
|
|
|