Skip to content

Latest commit

 

History

History
335 lines (288 loc) · 15.8 KB

TODO.md

File metadata and controls

335 lines (288 loc) · 15.8 KB

todo.md

this is my own scratchpad for keeping track of things. it gets truncated frequently.

see CHANGELOG.md for a more formal list of changes by release

done

  • remove gui1
    • done
    • remove original db-search
      • done

todo

  • gui, new tab for dedicated log
  • per-addon logging
    • I want the user to see a list of messages regarding that specific addon
    • when emitting a log line about a particular addon, capture that addon's source and source-id somehow
  • gui, download progress bar inside the grid ...?
    • pure fantasy?
    • defer until after job queue
  • greater parallelism
    • internal job queue
    • replace log at bottom of screen with a list of jobs being processed and how far along they are
      • each job can be cancelled/stopped/discarded
    • separate tab for log
      • that scrolls the other way
  • a 'stop' button to stop updates would be nice
  • download addon details in parallel
    • speed benefits, mostly
    • share a pool of connections between threads
      • N connections serving M threads
  • performance, check addons for updates immediately after loading
    • if after we've read the nfo data and we have everything we need, check the addon for updates immediately
      • don't wait for db loading and addon matching
        • we already have a match!
      • this might fit in with the greater-parallelism/queue based infrastructure
  • remove 'hostname' from log output
    • update privacy section in readme
  • bug, 'strongbox.version' in debug output is null when run as a binary

todo bucket (no particular order)

  • cli, exports

  • alpha/beta opt-in

    • user can opt to install alpha/beta/no-lib releases per-addon

  • import, export, capture 'pinned' information

    • we can now import addons at a specific version
      • when importing a pinned addon, should we keep the pin?
      • or drop the pin it and display updates?
    • we can now export addons at specific versions
      • I think we already have this though ... called :version

  • github, revisit the "-classic" suffix naming

    • this was solved more elegantly in curseforge-api

  • dedicated tab for "user-catalogue" ?

    • add, delete, update github addons
    • see accumulating release history for addons?

  • preferences, "update all addons automatically"

    • update README features

  • import, skip importing an addon if addon already exists in addon dir

  • import, why can't an export record be matched to the catalogue and then installed that way?

    • no need for padding and dummy dirnames then
    • installing normally would also include the mutual dependency handling

  • toc, add support for x-github key

  • create a parser for that shit markup that is preventing reconcilation

    • see aliases

  • if a match has been made and the addon installed using that match, and then the catalogue changes, addon should still be downloadable

    • right?
      • we have the source and source-id, even the group-id to some extent
    • switching catalogues may see the addon matched against another host
      • nothing wrong with that, but ...

  • http, add a timeout for requests

    • I have tukui API taking a looooong time

  • add support for finding addons by url for other hosts

    • wowinterface
    • curseforge
    • but these addons already exist in the main catalog ...
      • should they be taken to a search results page?
      • because what is presumably happening is the user can't find their addon in the search results (or can't be arsed to) and is saying "just install this please"
        • but wowman uses catalogs as a source of data, so if it can't find the addon in the catalog, then what?
          • fail? but the user just gave us a URL (UNIVERSAL RESOURCE LOCATOR) ! what is the fucking problem here?
    • the problem is expectations. wowman doesn't scrape addon host website HTML if it can avoid it
      • and user enters addon host website URL
    • this should be solved with more sophisticated catalogue searching
      • parse identifiers from URL, like source and source ID, then display search results
        • again, by encouraging the copying+pasting of URLs and then failing to find results when the URL IS RIGHT THERE AND WORKING we set ourselves up for failure and the user for disappointment/frustration
    • parking this
      • 2020-11-28: unparking this
      • since this was parked I've added source-id and source as standard across all addons
        • user just needs to copy and paste a url, strongbox matches it against catalogue, and installs it.

  • gui, feature, install addon from local zipfile

    • not the 'reinstallation' feature, but literally selecting a zipfile from somewhere and installing it
    • would be good for installing older versions of an addon?
    • would be good for installing addons from unsupported sources
      • wouldn't be able to update it however :(
      • I think I'll stick with supporting sources of addons
        • rather than enabling ad-hoc installation of unsupported addons

  • search, results not updated when catalogue is changed

  • gitlab as addon host

  • investigate warn-on-reflections

    • I think there may be some solid performance gains by turning this on
      • remember to profile first

  • EOL planning, robustness, only download/update the catalogue after an existing catalogue has been confirmed

    • github is down, wowman is erroring with a 500
    • failure to download a catalogue shouldn't prevent addons from being displayed
    • failure to contact a host shouldn't prevent addons on other hosts from working

  • EOL planning, bundle a catalogue with the installation

    • load it as a resource with static-slurp, like we do with the sql?
      • also compressed so it's tiny?
    • behind the scenes we download and load the full-catalogue
      • would this block reconciliation?
        • perhaps if there are unmatched addons after reconciliation we then wait and try again ...?

  • add checksum checks after downloading

    • curseforge have an md5 that can be used
      • unfortunately no checksum in api results
      • they do have a 'fileLength' and a 'fingerprint'
        • fingerprint is 9 digits and all decimal, so not a hex digest
    • wowinterface checksum is hidden behind a javascript tabber but still available
      • wowinterface do have a md5sum in results! score

  • test, can gui-diff and main/test be pushed back into the testing namespace and elided from release somehow?

  • EOL planning

    • I'm not going away and neither is strongbox, but! should I or my free time disappear will strongbox continue being useful?
      • what can I do to ensure it is the most useful if I just give up on it tomorrow?
        • catalogues have been pushed into user config, so they can be swapped out if necessary
          • more can be done around this to make catalogue generation more accessible and safe
      • what can't I control?
        • addon hosts
          • our interface with them is their API or in wowi's case, their API and website

  • reconciliation, add dirname support

    • not sure which hosts support these

  • wowinterface, multiple game tracks

    • investigate just what is being downloaded when a classic version of a wowi addon is downloaded
    • see 'LagBar'

  • bug, changing sort order during refresh doesn't reflect which addon is being updated

    • I think changing column ordering and moving columns should be disabled while updates happen
      • just freeze or disable them or something.

  • when curseforge api is down users get a wall of red error messages with very little useful information

    • see issue 91: ogri-la/strongbox#91
      • the error message has been improved but we still get a red wall of text
      • aggregate error messages?

  • reconciliation, rename 'reinstall all' to 'reconcile'

    • steal from the best
    • make the reconcile automatic
      • if a nfo file isn't found
    • remove the 'first time instructions' from the readme
      • it should just fucking do it

  • investigate better popularity metric than 'downloads'

    • if we make an effort to scrape everyday, we can generate this popularity graph ourselves

  • add a 'tabula rasa' option that wipes everything

    • cache, catalog, config, downloaded zip files

  • coloured warnings/errors on console output

    • when running with :debug on the wall of text is difficult to read
    • I'm thinking about switching away from timbre to something more traditional
      • he's not addressing tickets
      • it may have been simpler to use in 3.x.x but in 4.x.x it's gotten a bit archaic
      • I can't drop hostname without leaving pretty-printed stacktraces behind

  • cache, make caching opt-out and remove all those ugly binding calls

    • bind the value at core app start
    • this may not be possible.
      • binding happens at the thread level
      • if we start doing download concurrently, we need to pass our binds to the threads
        • which I'm not sure if is possible
      • moving back into bucket until I get around to doing parallel downloads

  • testing, capture metrics with an eye to improving performance and speed

    • we have coverage metrics now
    • would like some timing around certain operations
      • like loading the catalog
        • done
      • like downloading and installing the top-10, top-20, top-N addons
        • this could be a good benchmark actually
          • how quickly can one go from 'nothing installed' to '20 addons installed' ?
          • could be tied in with backups/exports
            • got to have backups+imports happening first
      • identify slow things and measure their improvement

  • gui 'wow' column is inconsistent

    • for curseforge, it's pulling it's value from :gameVersion, which may be empty
      • in which case it pulls it's value from the toc file, which may be different from the selected game track
    • since this is the 'installed addons pane', should the value reflect the value of the installed addon?
      • (and not the value of the addon to be installed)
      • and would this be inconsistent with the other fields that are also changing with new catalog information?

  • gui, toggleable columns as a menuitem

  • internationalisation?

    • Akitools has no english description but it does have a "Notes-zhCN" in the toc file that could be used
    • wowman was mentioned on a french forum the other day ..

  • gui, toggleable highlighers as a menuitem

    • highlight unmatched
    • highlight updates
    • touch of colour against each menuitem would serve as a legend

  • gui, have the log scroll the other direction

github

  • github, add support for user supplied github token
    • necessary if they want a large number of github addons without hassles
  • github, add a github catalogue
    • just a simple list of wow addons on github that can be installed with strongbox
  • github, importing an exported addon list with a github addon won't see that addon installed
    • unless that addon is present in the user catalogue
      • which in a fresh install where a list of addons are being restored is unlikely...
    • this is interesting actually. the exported addon list has become a mini-catalogue
      • some addons require the larger catalogue to resolve
      • github addons are resolved and installed by a different means...

import/export

  • export to markdown
    • I think I'd like a simple list like:
    • of course, this would be a different type of export than the one used for import
      • although ... I could possibly parse the list ... and nah.
    • clostache?

search

  • search, order by date only orders the current page of results
  • search, group results
    • group by downloads/age/category?
      • it would finally be the best use for category data

cli

  • cli, update specific addon
  • cli, install specific addon
  • cli, colours!
  • cli, replace with a repl
    • lein --cli gives you access to the code directly

unified UI

  • remove log split
  • remove tabs
  • gui, both panes, filter by categories
  • gui, group results
    • installed
    • updates
    • category ...

wontfix

  • add a 'Delete all' option to cache menu
    • we don't really want legitimate nfo files to be accidentally deleted
  • nightly unstable builds
    • building the 'develop' branch once a day
      • making it available as the 'unstable' release that always gets replaced
    • project.clj "x.y.z-unreleased" would be changed to "x.y.z-unstable"
    • development would happen mainly in feature branches
    • too much effort for what? more user reports? I don't have that sort of time
  • investigate .csv as a human-readable but more compact representation
    • might be able to save a MB on extraneous syntax
    • might be able to speed up parsing and loading
    • depends on profile task
      • update: profiling happened, it's not the json loading that is slow it was many other things.
      • reading the file and parsing the json is actually very quick, validation is slower but necessary
      • json is just more flexible all around than csv.
  • windows support
    • windows is just the worst, most awful dystopian software I've ever seen and it hurts my soul every time I try to use it
    • I just plain hate it, it epitomises the very opposite of what I stand for and I refuse to work on it ever again
  • addon 'detail' tab
    • link to curseforge
    • donation url
    • other addons by author ?
    • list the hidden/sub dependencies
    • too vague, too open ended, too much effort
      • just send them to the official addon page
  • .rar/.tar.gz addons
  • fallback to using :group-id (a uri) if curseforge.json is not available
    • low priority
    • curseforge.json will only ever be missing:
      • fresh install and
      • your network connection goes down, or
      • you're a victim of github's 99.999 uptime rating
    • wontfix because:
      • group-id is 'group-id' and not 'uri'
        • it may change in the future
      • there is a really really slim chance of this actually happening
        • I don't think it justifies the extra complexity tbh
    • 2020-11-28: I think this was about being able to download catalogues (curseforge.json) when github is down
      • if a catalogue is embedded then we can always fall back to that
      • see EOL planning
  • gui, search pane, clear search button
    • I don't think this is necessary anymore
  • gui, stateful buttons
    • don't allow enabled 'delete selected' buttons if nothing is selected
    • not going to coddle the user. deleting nothing will see nothing deleted.
  • cli, interactive interface when no specific action specified
    • you have N addons installed. what do you want to do? (list, update, update-all, delete) etc
    • this is a huge amount of hassle