psyinfra/tipue-search-lite

Serverless search for static websites

tipue-search-lite's Issues

examine commTerms list

Lists of common words (aka stop words) vary greatly by search engines.

I imagine that what is most appropriate depends on the content and size of the index. However, I suspect that the list we have is a bit too extensive.

It should also be possible for the consumer of TSL to override the contents of the list.

document configuration deviations from TipueSearch

The changes are getting to the point that TSL is no longer a drop-in replacement for TipueSearch.

We will continue to maintain strict compatibility with the index generated for Tipue. But many configuration options have been removed or behave somewhat differently.

These should be documented.

setup CI with GitHub Actions

This should run ESLint (#8) on every push to master and PR.

simplify search code

The search code is quite verbose. A faster (and more readable) solution would be to use JavaScript's regex abilities.

page context does not always display matching terms

I think there is a case sensitivity issue here.

For example, the search for "blue" displays the correct summary (containing "blue"), but "BLUE" does not (which I think shows text from the beginning of the page).

explore Template strings to generate HTML

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Template_literals

document configuration options

TSL should be configurable without editing source

The set var is not addressable by the user.

This connects to a question about how TSL should be invoked. Currently it runs by default and relies on global variables. If that approach continues, then set should become global and inhabit a TSL-specific namespace. However, this is generally considered a bad idea.

IMO, allowing defaults to be overwritten with invoking TSL makes most sense.

"-" syntax doesn't negate

TSL "supports" the - syntax, but all it does is skip the term (treat it like a common word) rather than removing pages from the results that match the term.

review variables and use `let` and `const` as appropriate

setup ESLint

ESLint can help with catching style and syntax errors.

A subset of rules should be selected, and committed to the repository as a .eslintrc.json file.

timer should stop later

The HTML is built, and then inserted as a block into the page.

One of the first things built is the duration info. The search has happened, but not the building of results. This skips over a fairly reasonable chunk of work.

Though less linear, I recommend building all other HTML first, then prepending the results found + search time information.

tab name is updated only once

The tab name s updated with the number of results, but only for the first search.

"forrest" has 7 results and leads to:

 (7) Tipue Search

Every subsequent search keeps the 7, rather than updating each time.

demo/index.html: multiple searches lead to different behavior between inm7 docs and demo example

demo/index.html does not reset the timer when searching two times in a row the same word.

This leads to the timer displaying search times like 1610.32 seconds.

handle special characters

When search contains

...

The whole text gets displayed in bold, including all results from all other search results.

[

opening square brackets lead directly to error.
(This is not always true because e.g. "[ ", with a space as part of term works, but the problem should still be handled).