If you try and search by director or writer and no results are found, the program crashes:

rank blah
Refine search options:
year
director
writer
Input "N" to skip.
director asdf
src/moviecat.nim(79) moviecat
src/parser.nim(72) decrypt_command
src/parser.nim(61) insert_movie
src/parser.nim(54) find_movie
src/parser.nim(42) refine_search
src/search.nim(71) extract_val
/usr/local/nim/lib/system/fatal.nim(54) sysFatal
Error: unhandled exception: index out of bounds, the container is empty [IndexDefect]
Error: execution of an external program failed:

I should handle this by falling back on No director/writer found, and then default to an unrefined search (so return all matches). This is the behaviour that year refinements follow, so we should be consistent (instead of crashing...)

This must follow from #3 Once quotation marks are no longer necessary it should be possible to search by both at once, since moviecat should be able to pick out everything automatically.

If you rank a movie, and then rerank the same movie, it should overwrite the old ranking with the new one, which it does. But when you are performing the ranking, it uses the old database, with the movie included in it. This means that reranking a movie right after ranking it will lead to a different sequence of questions.

In theory this also means that you could possibly have to rank the movie against itself.

Somewhat self explanatory.

Not sure what the best way to go about this is, storing the help strings in the binary itself or including them as external txt files that can also be compiled into a manual of sorts?

This is going to be a mild challenge in some ways. Search and ranking by director/writer can be done the same way that it is done by release year. The challenge is that I will need to update imdb.db using the title.crew.tsv file.

Here's what needs to be done:

• Workflow for updating the movie database from input files on startup (to add new movies not in db)
• Workflow to import title.crew.tsv to add director/writer information to the database.
• Search by director/writer
• print by director/writer

Right now the ranking algorithm uses a straight binary search and insertion which is great from an efficiency standpoint but bad from a movie ranking standpoint.

In general I find that people's movie ratings tend to "degrade" over time. That is, people send to move towards either more neutral or more extreme opinions about movie quality the longer away it is from the date they watched the movie. This can lead to what I have discovered in myself as a natural movie inflation. I rarely consider a movie "worse" than the very first question, which is always dead center. This means that I seem to think every movie I watch is above average!

I propose a smarter algorithm to compensate for this effect:

1. First initial question comparing the to-rank movie with the most recently watched (if watch dates are input) or most recently ranked movie. Worst case effect on ranking: one additional question. Best case effect: reduce the number of questions to 1.
2. Choose comparison movies using some sort of "similarity" metric. For example, if selecting the middle of the range, select a movie "off-center" if its content is more directly comparable to the input movie. How to define "similar" is an open question, but this should lead to more fair and more accurate personal assessments.

When ranking a movie, each questions prints its question number out of the total number of questions. As evidenced by this example, the estimated total number of questions is wrong:

(1/10) Is (tt12718300) Confess, Fletch, 2022 > (tt15679400) Knock at the Cabin, 2023?
n
(2/10) Is (tt12718300) Confess, Fletch, 2022 > (tt14814040) Not Okay, 2022?
y
(3/10) Is (tt12718300) Confess, Fletch, 2022 > (tt8478602) Godzilla: The Planet Eater, 2018?
n
(4/10) Is (tt12718300) Confess, Fletch, 2022 > (tt0103644) Alien³, 1992?
y
(5/10) Is (tt12718300) Confess, Fletch, 2022 > (tt4244994) The Last Duel, 2021?
y
(6/10) Is (tt12718300) Confess, Fletch, 2022 > (tt6217926) Annette, 2021?
n
(7/10) Is (tt12718300) Confess, Fletch, 2022 > (tt0103639) Aladdin, 1992?
y
(8/10) Is (tt12718300) Confess, Fletch, 2022 > (tt0133240) Treasure Planet, 2002?
n
(9/10) Is (tt12718300) Confess, Fletch, 2022 > (tt2381249) Mission: Impossible - Rogue Nation, 2015?
n
What date did you watch this movie? (YYYY-MM-DD)
Input "N" to skip.
2023-03-20

It's probably an off by one error. I suspect that I estimate the number as (n_ranked div 2) + 1 instead of (n_ranked + 1) div 2.

Moviecat is relatively old-ish now, a few years at least, and frankly Nim isn't at the level of stability that I'd like for this software. The fact that the sqlite bindings were removed from the stdlib since starting this project in Nim 2.0 should be evidence of that enough, but the longer I remain in nim the older the zip bindings get and the more likely it is that they'll break.

All of this to say, I've been considering and am now leaning towards a full rewrite of moviecat in another language. I'm not sure which yet, but I have been considering Go, Swift, or even just going to Python. I'm leaning away from Python though because I'd like to one day build a proper app, and a lower level language will make that significantly easier.

When searching, for example when looking for ranking or when looking for stats, director and writer fields require surrounding the name with quotation marks. Movies, however, do not need to be surrounded by titles. We should move to a unified scheme for search strings.