Find who teaches your classes before everyone else.
Live: https://enesyilmazcode.github.io/Finder/
Ohio State's class search will confirm that CSE 2221 exists. It will not tell you which of its sections still has a seat, or which instructor people warn you about. Finder answers both. Search a course and you get every section grouped by instructor, with a rating next to each name and real seat counts where they exist.
osucoursesearch.org did roughly this job and went down. Its database stopped answering and its author moved on, so the site now returns a Django error page instead of course data.
Finder has no server and no database behind its search, so there is no equivalent piece to fail. It is a folder of files on GitHub Pages.
- Filters. Days, time window, busy times, minimum rating, rated only, hide full, hide online, hide permission-only, undergraduate only. A sort control sits above them.
- Middle. The same sections as a list by instructor, or on a week grid.
- Right pane. Rating, difficulty, take-again, seats, room, description.
The grid is the same sections placed by when they meet, and whichever one you picked stays open beside it.
On a phone the filters move behind a Filters button and the right pane takes over the screen, so you get the same three panes one at a time.
No filter touches the network, and nothing a filter hides vanishes quietly: the page says how many sections it removed and offers a button that shows them anyway. A separate Schedule view lets you keep sections from different searches, see them together on the week grid and catch conflicts before you register. Linked labs and recitations come with their section; when several are valid, Finder asks which one instead of guessing. The schedule stays in this browser unless you choose Share.
Every search is in the address bar, and so is the section you picked, so a link opens on the section you meant rather than on the search. The right pane copies the class number BuckeyeLink asks for, and offers Share on the browsers that have it.
flowchart TB
accTitle: Where Finder's data comes from
accDescr: RateMyProfessors and Barrett's schedule both refuse to answer a web page, so a job on GitHub reads them every morning and commits each answer to this repo as a file the site serves. Ohio State's own class API answers a web page directly, so your browser asks it the moment you search.
subgraph nightly ["Earlier today, on a computer at GitHub"]
direction TB
rmp["ratemyprofessors.com<br>refuses a web page"] --> ratings["data/ratings.json<br>every rated OSU professor"]
ratings ~~~ barrett["asc.ohio-state.edu<br>refuses a web page"]
barrett --> seats["data/seats-1268.json<br>Autumn 2026 seat counts"]
end
ratings --> committed["committed to the repo,<br>served like any other file"]
seats --> committed
subgraph browser ["Right now, in your browser"]
direction TB
typed["you search CSE 2221"] --> osu["content.osu.edu/v2<br>answers a web page,<br>every section seconds old"]
end
committed --> page["the page in front of you"]
osu --> page
data/ratings-courses.json rides along with the ratings, and
data/courses.json is rebuilt the same way once a week.
Only the saved files can go stale, which is why the page prints the date the seat counts came from.
The split is not a preference. A website can refuse to be read by a page
from another website, and RateMyProfessors and Barrett's schedule both do.
The Ohio State address in that picture does not. It sends one header,
Access-Control-Allow-Origin: *, which is standing permission for any page
to read it. That header is why Finder can exist with no server.
Barrett's file also carries a compact instructor listing, and it can land there before the primary instructor appears in Ohio State's class API. The nightly snapshot keeps that fourth field beside the seat counts. Finder uses it only when the live response has no primary instructor and gives the live Ohio State assignment precedence as soon as it appears.
OSU's class API returns one enrollment number per course and stamps it onto every section. It publishes no per-section enrollment cap at all. For CSE 2221 in Autumn 2026 it calls all 22 sections open. 12 of them are full. Class 4831 is 41 people in a 40-seat room with two waiting, still listed open.
Seats come from Barrett's schedule instead, a fixed-width text file that
asc.ohio-state.edu rebuilds once a day around 06:50 Eastern. Barrett
lists fewer sections than the API, 407 of Autumn 2026's 1,064 CSE sections
when I checked on August 20, so some rows show no seat numbers. That beats
printing a zero nobody checked.
About 36% are, measured across 757 instructors in 8 subjects for Autumn 2026. Unrated is the ordinary case, so it costs nothing on screen. Every name is a link either way, to a profile when there is a match and to a RateMyProfessors search when there is not. Matching drops middle names and suffixes, so OSU's "Diana Ikenberry Kline" finds RateMyProfessors' "Diana Kline". When two professors share a first and last name the lookup returns nothing rather than picking one, because a wrong rating is worse than no rating.
A 3.0 can be a pile of threes or two camps that never met, and it can be earned in a course nobody on screen is taking. The detail pane draws the five per-score counts as a bar, and says how many of the ratings name the course you are looking at. Paolo Bucci, pulled on August 21: his 3.0 is 32 ones and 38 fives, and 52 of his 147 ratings are for CSE 2221.
Those course names are free text raters typed, so "CSE 2221", "cse2221", "CS2221" and "COMPUTERSCIENCE2221" all count, and so does a bare "2221" unless the same professor also wrote that number under another subject, in which case it is too ambiguous to give to either. The pre-semester CSE 321 never counts, because nothing says the old number is the new course. OSU writes 23% of its catalog as "1110.01", which no rater types, so those match on the 1110 and the line says 1110 back.
The counts live in data/ratings-courses.json, 151 KB gzipped and read
only by the detail pane, so it is fetched when you open the first section
rather than when the page loads.
A rating is what people felt. data/grades.json is what Ohio State wrote
down: five academic years of per-section grade distributions, Autumn 2021
through Spring 2026, obtained under Ohio's public records law because OSU
publishes nothing like it. scripts/fetch-grades.mjs folds that spreadsheet
into one curve per instructor per course.
This is the one join in Finder that does not have to guess. RateMyProfessors
has to be matched on a name, which is why two professors called Alan Reed get
no rating at all. The records request asked for the instructor's OSU address,
which content.osu.edu already publishes beside every section, so a curve is
keyed on bucci.2 and lands on one person or on nobody.
The pane leads with the average and the share who got an A or A-, because the two disagree exactly where it matters. A section that is half A and half E averages the same as a section that is entirely B, and only one of them is a coin flip.
Withdrawals sit beside the curve rather than inside it. Nobody who withdrew has a grade, so they cannot move a mean, which means a course a third of the class drops can post a flattering one. The number that explains it has to be on screen next to it.
The same professor teaching the same course to honours students in Autumn and to a summer section averages two different curves, and neither is their "difficulty". Finder prints how many sections and how many terms went into each one so the number carries its own sample size, and says so outright under thirty students. It cannot tell you who the students were, and that is usually the larger half of the answer.
Sections too small to publish are missing on purpose. A distribution over four students can identify one of them, so Ohio State withholds those, and the request asked for the withheld rows to be marked rather than dropped. That is why the pane can say "two more sections were too small to publish" instead of quietly showing a curve with a hole in it. A course where every section was withheld says that too, rather than looking like a professor with no record.
data/grades.json is absent until the request is fulfilled. The loader reads
a 404 as "not published yet" rather than as a failure, so the block simply does
not appear, and nothing else on the page changes.
flowchart TB
accTitle: What one search does
accDescr: A search asks Ohio State for the sections and reads the two snapshots that started downloading when the page opened. rank.js merges and scores them, render.js groups them by instructor, and changing a filter redraws from memory without touching the network.
typed["you type CSE 2221 and press Enter"] --> osu["content.osu.edu/v2<br>the 22 sections, live"]
typed --> snaps["data/ratings.json, Paolo Bucci 3.0 from 147<br>data/seats-1268.json, 5168 has 33 of 40"]
osu --> rank["rank.js<br>merge the pages, score every course against your words"]
snaps --> rank
rank --> render["render.js<br>group the sections by instructor, rating beside the name"]
render --> page["the page you are looking at"]
page -- "change a filter" --> filters["filters.js<br>redraws from memory, no network"]
filters --> page
The two files start downloading when the page opens, so a search usually waits on Ohio State alone.
Only the newest search is allowed to draw, so a slow first search can never paint over a fast second one.
That diagram is only the part a search waits on. Opening the page is 30 requests
across two hosts, counted in Chrome against a local copy with the cache
cleared: 29 files from this repo and the term list from content.osu.edu. The
29 are the HTML, two stylesheets, three font files, nineteen modules, the
ratings snapshot, two seat files and the favicon. The live page adds one more,
the page-view ping to the analytics worker, which local runs skip.
Opening a section adds requests the page load does not make: the rating course
codes, the grade curves, the list of instructors who have a photo, and, if this
one does, their headshot from opic.osu.edu. That is Ohio State's own photo service and a third
host, and the request tells it which instructor was opened.
The share card is not one of them. og.png is fetched by whatever is unfurling
the link and apple-touch-icon.png by iOS when someone saves the site to a home
screen, so neither costs a visitor anything.
None of them go to Google. The three families are woff2 files in
assets/fonts/, served with everything else.
No build step and no dependencies. Serve the folder over HTTP:
python -m http.server 8000Then open http://localhost:8000. Opening index.html off the filesystem
fails, because file:// pages cannot make cross-origin requests.
The snapshot scripts are Node 22:
node scripts/fetch-seats.mjs 1268All four refuse to write a snapshot that came back far short of the one already
committed, and leave the old file in place. FORCE_WRITE=1, or the force input
on the workflow, writes it anyway, which is how a real shrink gets shipped. A
file that failed to parse is refused either way, and so is a short answer from
the term list, which takes ALLOW_TERM_DROP=1 because it deletes the files of
terms the run never fetched.
npm test runs the suite through node --test, with nothing installed. The one
file that calls Ohio State is skipped unless FINDER_LIVE=1, so a default run
stays offline. The runner prints the count at the end; it is not written out
here, because the copy went stale every time the suite grew.
index.html the whole page
css/finder.css the app stylesheet, no framework
css/fonts.css the three families, self hosted
css/stats.css the /stats page
assets/fonts/ the woff2 files and their licenses
js/
app.js wiring, state, URL sync
api.js client for OSU's class API
rank.js merging, scoring, splitting results
render.js list view, grouped by instructor
calendar.js week grid
schedule.js saved sections, sharing and conflict detection
detail.js right pane
filters.js client-side filtering
sort.js ordering by rating, difficulty, seats or start time
deeplink.js the section a link points at
ratings.js RMP snapshot and name matching
grades.js five years of grade curves, joined by OSU address
headshots.js which instructors have a photo, and the initials fallback
seats.js Barrett snapshot and the term guard
trend.js what moved between two seat snapshots
courses.js lazily loaded course index
format.js days, times, units, names
hit.js one page-view ping, skipped on localhost
analytics.js the one place the worker URL lives
stats.js the /stats dashboard
scripts/ the four snapshot jobs, and the records import
data/ the snapshots, committed
docs/ what OSU's API, Barrett's schedule and the records get wrong
tests/ node:test, zero dependencies
wireframes/ three layouts considered first
analytics/ the page-view counter, a Cloudflare Worker
stats/ the page that reads the counter
- Seats are a morning snapshot, so a section can fill before you search.
- Barrett covers fewer sections than the API, so some rows never show seats.
- Back closes the section pane and puts back the filters it opened with, but it does not undo a search.
- A shared schedule refreshes each section separately, so a long schedule makes several requests to Ohio State when it opens.
- Changing a filter clears whichever section you had selected.
- Columbus only, and only the three terms OSU's API exposes at a time.
- A section listed in two rooms of the same building at the same hour shows one of them.
Finder is not affiliated with or endorsed by The Ohio State University.
MIT. See LICENSE.