ijkilchenko / fuzbal Goto Github PK

Chrome extension: Gives Ctrl+F like find results which include non-exact (fuzzy) matches using string edit-distance and GloVe/Word2Vec. Also searches by regular expressions.

Home Page: https://chrome.google.com/webstore/detail/fuzbal/lidjpicdkcgjdkgifmmpalkibjeppdof

License: MIT License

JavaScript 83.08% CSS 2.23% HTML 14.69%

fuzbal's People

Contributors

Stargazers

Watchers

Forkers

adrianloer matthewc-mps-aust lkhamsurenl-zz hbcbh1999 truong225 goyalakshat14 rahimacm tangert ssimono chrome-extension eugenesvk

fuzbal's Issues

Start searching for text when typing outside of a text field (without having to open the extension)

This is a suggestion by a friend to mimic how Firefox lets users find results. That is, Firefox lets users just type without first doing Ctrl+F and shows results (briefly) if a user has not selected a text field.

At first thought, this is not an enhancement I will implement because that's not how Ctrl+F behaves in Chrome and mimicing that functionality seems to be more important right now.

Order results from top to bottom

The results are ranked by their Liechtenstein distance but, somehow, even if two results are exactly the same, right now they might be out of order with respect to the order they appear on the page. This is a bug.

Use JQuery better

Do $('#helpTips').hide() instead of $(document.getElementById("helpTips")).hide();.

PDFs are not supported.

Unfortunately, I have no idea how to parse the text of PDFS. This would have been good to be able to do, but it may become a wontfix issue.

Investigate how to fix it in any case and investigate what other formats are not supported.

JSON data not necessarily loaded

Initially this extension didn't work, and there were no error messages in the popup.
Later I checked the background page and found these errors:

_generated_background_page.html:1 Error in event handler for (unknown): TypeError: Cannot use 'in' operator to search for 'skip' in undefined
    at subset (chrome-extension://lidjpicdkcgjdkgifmmpalkibjeppdof/background.js:20:18)
    at chrome-extension://lidjpicdkcgjdkgifmmpalkibjeppdof/background.js:28:24
3_generated_background_page.html:1 Error in event handler for (unknown): TypeError: Cannot use 'in' operator to search for 'menu' in undefined
    at subset (chrome-extension://lidjpicdkcgjdkgifmmpalkibjeppdof/background.js:20:18)
    at chrome-extension://lidjpicdkcgjdkgifmmpalkibjeppdof/background.js:28:24

subset(msg.words, vectors) fails if vectors is not yet loaded.

Add a loading icon

I want to add a loading icon that will appear while the results are being processed.

Develop a README

Can't find Fraunce when looking for France.

Website: http://www.luminarium.org/renascence-editions/stella.html

Keyword: "France"

Expected result: It finds the weird spelling "Fraunce"

Actual result: It doesn't find it

Note to self: use Set() for uniqueness operations

Support double quotes around multiple words

Right now if there are double quotes around something, it is assumed it is only one word, but we need to stop the expansion of anything inside the double quotes.

Consider not changing the order of find results when results are cycled through (or clicked on)

It might be a bad design decision to keep the active find result on the top all the time and change the order in which the find results are shown.

When looking up vectors, wait until vectors are returned

Consider not getting rid of numbers

Shouldn't be hard to fix.

Possible interference with a website's buttons

The top navigation buttons don't work with the extension turned on and work with the extension turned off on this website: https://us.pycon.org/2016/

Consider adding wildcard or full regular expressions support

We might consider letting a user use * which will match, say, at most 5 words.

We could also support full regular expressions and let the user type something like /foo..bar/ into the search box to search for a specific regular expression.

Feature request: Keep search box open, and remember place on search results

Hi there, thanks for the great extension!

You open the search box, type a term, and jump to one of the results. If you close the box to read the webpage, you have to retype in the search box again. Just wondering if there's a way to keep the box on top, and keep your place in the search results.

Best,
Jeff

how do i index more text?

how do i index more text?
The text in (some) of my webpage is NOT visually appearing in your match results UI (as-in fuzbal works on some of my page's text but not all of it. It shows matches for stuff in the header but not in a particular <a tag </a>

Troubleshooting performed: I have refreshed the page after activating the chrome extension, no luck.
Any existing approach to indexing more page text so that fuzbal considers more matches?

Here is the snippet of HTML on my page that Fuzbal DOES NOT seem to handle (the text that I expect it to handle is literally this camel-cased:

TextThatFuzbalDoesNotShowInItsMatchesUI

<a class="alps-anchor alps-grid-col-0" href="/a1234556256" tabindex="-1" id="123_anchor" title="TextThatFuzbalDoesNotShowInItsMatchesUI "><i class="alps-icon alps-checkbox" role="presentation"></i><i class="abc" role="presentation"></i>yeeee </a>

Try using word2vec output trained over a large data set

Current version, v0.91, uses the output from word2vec which is generated from their default data set (text file of 100 mbs). This output gives a json file which is about 150 mbs (and the extension is then compressed to about 55 mbs in the Chrome Store). This json file takes up about 400 mbs of RAM when the background page reads the file into a Javascript object (for instant vector lookup).

For comparison, 400 mbs of RAM is 4x bigger than what Adblock uses in RAM and 8x bigger than what Lastpass uses. So reducing the word2vec output file somehow would be great.

On the other hand, we need to experiment with training over large data sets and seeing how large the output files become. The added accuracy of word vectors might be very useful.

Try some of the following things while keeping in mind the size/accuracy trade-offs.

Train over several gbs worth of training material
Cut the number of dimensions in half (should save about half the space?)
Round up the floating point values a few digits
Somehow only keep most frequent words in the json format
Split the json file into two files: one with 10% most frequent words and one with the other 90%
- Allow the user a setting to choose to load the larger file into RAM as well

Sometimes, the active result is not the one actually in green on the page

There are some inconsistencies with which ones are active (possibly in the cases when there are more than a hundred results total).

Tell mac users to use "command" instead of "ctrl"

Also possibly mention "command" in the various descriptions so as not to alienate the delicate mac users.

Add the ability to search for words that are not next to each other

Eg.:

Text on page:

this is a test about search

Search string:

this is a search

Both "this is a" and "search" should be highlighted

Suggestions for unit testing and integration testing

Right now there is no automated tests that can be run, unfortunately. I am not sure how best to perform testing of a Chrome extension (or a JavaScript project in general since I am not yet familiar with JavaScript).

One thing that I would like to do is set up some JavaScript files which can be run with NodeJS. I could refactor some of the methods that I use in content_script.js into a library and then test those methods separately from various Chrome extension specific methods/variables.

Another thing that I would like to explore is mocking Chrome methods/variables using some sort of a library like https://github.com/acvetkov/sinon-chrome but the usefulness of this is unclear at this point. Any suggestions are welcomed.