ePub text search

[dghag]

How to implement "ePub text search" feature?

[dghag]

Please anyone at least tell me, Is "text search" possible or not??

[fchasen]

For the immediate future search isn't possible, though we are planning to develop a backend to enable it.

It will likely be separate from the ePub.js code, as it will require a python.

[dkurth]

I wonder if you could build the search indexes using lunr.js. It already supports tokenization, stemming, and stop words. I haven't used it for anything serious, so I don't know about performance, and building the indexes on the client is not optimal, so just a thought.

[fchasen]

@AJRenold - could you update this thread with the search progress and close?

[AJRenold]

We're developing a light-weight indexing and search tool which allows an entire epub file to be searched via a search API. (https://github.com/futurepress/epubjs-search) The reader plugin search.js (https://github.com/futurepress/epub.js/blob/master/reader/plugins/search.js#L4) points to the API endpoint where you must be running a search server with the index of the same book. This model only supports indexing and searching a single book.

I'll update the epubjs-search readme today with better instructions on how to use the library.

[grayxr]

Please update the epubjs-search readme. Thanks!
@AJRenold @fchasen

By the way, would a nice alternative option be to just use the unzipped/decompressed epub folder format and perform a search query that iterates over all of the epub's '.xhtml' files?

[AJRenold]

I just updated the epubjs-search README.md with instructions that should get you up and running a simple Python search API.

We are actively considering other options to create a search feature that runs on the client side only.

[fchasen]

The search engine needs to have a sense of how a epub is put together or it can't generate the Epub CFI and cfi's are what allow the reader to jump to a location in the book.

To briefly describe what the search engine does:

• Parses the epub spine / manifest
• Indexs each page that was in the spine along with the spine items cfi position
• Retrieves pages from the index that match a search query
• Looks for all the instances in that page of the query and assigns them full cfi's
• Returns the list of results as json

[shenzhuxi]

Is lunr.js good for this issue?