stscoundrel / cleasby-vigfusson-next Goto Github PK

• Use current year
• Mock for tests to keep deterministic

Some definitions (probably incorretly) may have only one letter, rest of the definition being parsed into second definition,

While it matters little in the word definition page, it does not look great as the selected definition in the frontpage. Just add min length to the randomizer algorithm, producing entries that always have content

Some headwords are written in allcaps, while some are capitalized and some are lower case.

This generally does not matter much and does reflect the original source material, but the uppercase ones seem a bit odd in middle of sentences.

Either:

• Lowercase them in word pages in middle of sentences
• Or lowercase them in dictionary service already

SEO, Schema & Link use example data from Dictionary dependency. While nice idea to use real data, whenever the dep changes the tests have to change too.

Copy / hardcode dictionary data that is to be used for those tests.

• Back to top scroller
• Only visible in mobile
• Add unit + integration test cases

Integration tests build the whole thing, eg. ensure build works. No need to do it twice in GH actions.

Should request fail (=timeout, most likely), gather failed batches into retries. Run them with smaller batch size, like 100.

Word link just translates the word to a slug without making it lowercase. Should probably just use the 'slug' property from the word.

Could run cleasby-vigfusson-abbreviations through abbreviatrix to autoadd abbr tags to descriptions. See if makes sense.

Lighthouse seems to punish for large list of DOM nodes

Would be better to display content when searching

Should provide alternative avenue for building all ISR pages.

Current problem: many more pages that can be build & deployed to Vercel. There is hard limit of 16 00 files. Once the project is deployed, it can generate as many more as it needs. But it has to happen after deployment.
Current solution: use ISR to generate pages as users open them. Downside is that the first load can then be slow. Other approach is to use Gangleri, a tool to walk sitemaps, therefore triggering rebuilds. However, that is slow as it makes separate request for each and can be spammy.

New solution:

• Add API endpoint
• Protect with env secret. Important as hell, as this could potentially cause a lot of processing spam.
• Trigger revalidate of either all pages, or a subset based on some param. See how long rebuild takes etc.

Info text for runes tends to put the rune at the end of screen. Often gets broken into next line, which hurts reading

Letterlist in mobile is kinda awkward. Hide behind hamburger.

Currently the "random entries" section in frontpage feels a bit weird. Some things that could improve it:

• Smaller amount of entries (less screen space)
• Alphabetical order
• Possibly definition excerpt? May need to limit entries to those that actually have one definition of reasonable length. Considering special nature of Cleasby & Vigfusson, may not be possible. Would probably work better with Zoegas shortform definitions

Should ignore casing when looking for results.

SWC seems to now support all Babel features used in the dictionary.

• Switch to SWC compile
• Switch to SWC minimize
• Drop Babel deps
• Drop Terser dep. Probably not needed in current version

Add it under "dictionary project". Originally made for this particular use case

Some json-dl contains strings that are not present in content, make adjustements

When searching from main bar, criteria is unfilled

Currently can output same abbreviation many times, if it appears in different definitions

Using younger futhark, add possible runic inscriptions of headwords. Just transform letter by letter. Would vary from historical forms, but interesting

Cleasby & Vigfusson uses a lot of abbreviations. Scan definitions & output list of used appreviations at the end of page.

At least to single word page

O caudata (= ǫ) would be more accurate letter for words that use "ö" in the Cleasby Vigfusson book. Using "ö" in Old Norse is a more modern practice that is mostly based on modern Icelandic and/or lack of letter ǫ in all keyboard setups.

Should the word contain letter ö, offer alternative version with ǫ. When that is done, also offer an explanation as to why this alternative form exists. Similar work is done for Old Swedish dictionary in this PR.

Allow searching by headword or definitions

Currently generates on demand each time, but contents of the file haven't really changed from the start. Just use .txt file in public & drop the page path

Folks seem to search "english old norse dictionary" and the likes. An oversight in SEO is that frontpage wont mention what languages this dictionary maps.

Open it up for easier google results.

The Cleasby & Vigfusson title will still exist in sitename and all over content, but looks like the language itself would work better in SEO titles.

Nowadays Next.js supports having "src" folder instead of top level pages. Simplify top level by adding src folder

Similar to Old Swedish dictionary.

• Add page for listing all works, authors and sources of the book
• Should there be works & authors cited, add a link to the all sources page

Similarly to how Old Norwegian dictionary (a sibling project) handles that. Out of context random runic inscription might not be too helpful.

Individual word page does not say where the definition is from, ie. the Cleasby/Vigfusson Dictionary. Add something like:

As defined by the Cleasby & Vigfusson Old Norse to English dictionary:

Schema description has tags depending on content. Filter out html.

Not a full rewrite, but deps & defs that let JS & TS play together. Much like Zoega version