Degree requirements and observed paths

Read the Degree Explorer rules before changing the feature. That file records the product decisions and prevents interface drift.

Institution files are registered in CEDAR_CONFIG_FILE_SPECS and validated by R/lists/degree_catalog.R at startup:

Table	Grain and purpose
`degree_catalogs.csv`	Snapshot: program name, degree label, year, source, retrieval date, coverage and caveats
`degree_rules.csv`	Rule node, parent, catalog, phase, common area, target course, operator, minimum, grade, review status and detail
`degree_rule_courses.csv`	One course per leaf rule, with catalog title and credits
`degree_roadmap_nodes.csv`	Authored planning slots, course/group labels, explicit lower/upper course level, credits, type, paired flag, source and `row_group` course-family alignment; `*` holds shared entry routes
`degree_roadmap_links.csv`	Selected prerequisite route edges, earlier vs earlier-or-concurrent timing, detail, source and evidence ID
`degree_roadmap_courses.csv`	Exact course components for concrete tiles; two rows for lecture/lab bundles, none for unresolved choices; no historical equivalences
`degree_roadmap_options.csv`	Finite OR sets: `option_id` groups required AND components within an alternative; minimum option count, course code/title, component credit range, individual course URL, program evidence and optional course-prerequisite evidence. Separate from selected bundle components
`degree_catalog_evidence.csv`	Exact excerpt, catalog year, program/course page, section path, retrieval date, transcription/review/conflict status and explanatory note
`degree_roadmap_basis.csv`	One program/node membership: degree, prerequisite preparation or illustrative general route, with a citation and explanation
`degree_roadmap_plans.csv`	Degree credit target, usual/maximum loads, baseline semesters, first upper-level semester, usual allowance-slot credits, Gen Ed source URL, target citation and workload assumption
`degree_roadmap_starts.csv`	Common math/English placement identifiers, English course routes and explicitly completed degree credits
`degree_roadmap_math_routes.csv`	One route per program/math placement identifier, with program-specific labels and ordered shared course IDs; no calculus fallback
`degree_roadmap_conditions.csv`	Course-count pools and permission conditions, target, minimum, eligible single-course slots, visible label and course-page evidence; separate from named-course edges
`pathway_subject_groups.csv`	One descriptive group per subject prefix, with consistent labels

Degree tables are exposed as CEDAR_DEGREE_CATALOG (catalogs, rules, courses, evidence); templates as CEDAR_DEGREE_ROADMAPS (nodes, links, starts, basis, plans, courses); categories as CEDAR_PATHWAY_SUBJECT_GROUPS. The synthetic institution has empty catalog files rather than importing real university rules as synthetic facts; its fixture courses still have descriptive groups.

Readable rules behind a plan

build_degree_plan_rules(catalog, roadmaps, view) in R/features/degree-explorer.R turns the actual selected view and saved metadata into a readable table: recorded catalog outlines, displayed courses/groups, finite alternatives, prerequisite edges, aggregate/permission conditions, credit budget, authored semester positions and other preferences. Underlying node IDs remain attached when identical elective slots share a row. Course-count conditions never become named prerequisite arrows. Catalog rows retain captured wording and links; CEDAR preferences have no invented catalog quote. Recorded admission/GPA rules are not student eligibility checks.

degreePlanRulesUI/Server(id, catalog, roadmaps) mounts as Admin Catalog Review’s default Visualization rules view. Degree Explorer’s Rules behind this plan dialog uses the same payload/table, as does the CSV download. This explanation requires no individual approvals or independent parser/AI proposal.

The health pilot adds Population Health, Medical Laboratory Sciences, Nursing (Freshman Direct Entry) and Radiologic Sciences (Nuclear Medicine). These are catalog-derived authored excerpts in the existing institution CSVs. Clinical admission, capacity and cohort calendars are not evaluated. Population Health’s two broken entries use explicitly provisional 3-credit slots, not invented identities/official credits. Published total conflicts stay visible.

kind = cohort stores a required block with every component in degree_roadmap_courses.csv: at least two components, possibly mixed levels, upper-level if any component is upper-level. The complete block stays together under the credit cap. component_count makes chain metrics count five courses in a five-course block rather than treating every block as a two-course pair. This does not imply every component is a prerequisite of every later course; link details and quoted sources describe the selected grouped progression.

Rule semantics

The reference sets under institution/<id>/shared_requirements/<year>/ separate University, Gen Ed and awarding-college policies from degree-specific trees. They reuse the operator vocabulary and add scope, overlapping-credit roles, bounds, source pointers and attention cases. read_shared_requirement_bundle() validates their data; an explicit capture directory additionally verifies source bytes and locators. format_shared_requirement_preview() produces the readable rulebook. capture_shared_requirement_sources() makes pinned-edition source candidates without modifying rules. The offline entry point is scripts/shared-requirements.R; the institution README documents the format and annual workflow. These files now supply illustrative semester allowance allocation and upper-credit context; they are not a transcript evaluator.

all and choose_n preserve AND/OR trees. A two-course all child is one option of its parent’s choose_n. Course leaves distinguish course, prior, concurrent, and prior_or_concurrent. credits records a threshold; eligibility restrictions remain explicit until encoded. condition preserves non-course rules without pretending they are executable. none is a known absence; missing rows mean not encoded. Bad references, cycles, impossible choices, missing leaf courses and duplicate grouping keys fail loudly.

Admission and degree rules share structure without merging their purposes. Requisites have a target course and separate phase. No evaluator yet determines student eligibility, admission or completion. Do not infer course equivalences or impose the current catalog on historical students. A future evaluator needs allocation constraints: one course may satisfy permitted overlapping rules, while earned credit counts only once. Coverage and review must gate planning use.

Semester planning

build_degree_roadmap(catalog, roadmaps, catalog_id, math_start, english_start) selects one program plus its entry routes. build_degree_roadmap_view() adds presentation text; degree_semester_board() (R/modules/degree-explorer.R) renders the same payload as an HTML semester board. www/cedar-roadmap.js draws connections and handles hover and focus; it makes no scheduling decisions and does not walk the graph: get_degree_chain_metrics() exports each course’s upstream_ids and downstream_ids, and qualifying_ids names a course-count condition’s pool. format_degree_explorer_preview() formats that production payload for analysts without Shiny.

After scheduling, assign_degree_roadmap_rows(nodes) adds a positive integer row from the authored row_group family. Math and writing have reserved rows; parallel courses occupy separate rows. Other families can share a row only outside each other’s complete semester span, preserving gaps within sequences. The planner shifts these rows together below the early Gen Ed allowance band. This step changes neither course semesters nor workload. The view adds n_rows; the renderer uses shared CSS grid tracks so each row grows to fit its tallest tile across every semester. Browser code still only draws actual graph edges.

Nodes distinguish major, support, writing, preparation and course choice. Paired lecture/lab coursework is a single node with combined credits. This avoids misrepresenting same-term bundles as a prerequisite cycle. Each edge is one selected valid prerequisite route, not the complete AND/OR rule tree. Missing shared nodes are satisfied by the selected placement scenario. Biology currently illustrates Calculus I rather than its alternative Applications of Calculus.

Finite alternative sets live in degree_roadmap_options.csv, not the component table. All five Philosophy sets are transcribed completely. Their nodes are unresolved choices, with the minimum/count and a three-line course-code preview; the popup lists every title, credit range, level and course URL. The program evidence must match the node’s basis. Optional prerequisite evidence must match that individual course page and catalog year. Selected-course edges/conditions on an unresolved set are rejected. Logic reserves three credits for a lower-level choice and explicitly identifies the higher-level, variable-credit alternative. Offering history for a finite set uses its options separately in the standard table; it never aggregates their seats as a required bundle.

Rows sharing (catalog_id, node_id, option_id) form one complete alternative. Validation sums component credit ranges before checking the planning allowance; option counts count groups, not rows. The feature supplies typed option groups for compact card previews and grouped popup lists, while history retains exact individual course identifiers. requirement_summary on a node is authored institution text for a shared block’s total/rules, distinct from slot credits. CS’s science block has two four-credit lecture/lab slots and two three-credit science slots. The first two enumerate eleven complete alternatives (nineteen individual lecture, lab and combined-course entries); the other two list the eleven lecture/combined courses. Higher-credit choices and the catalog’s open advanced-science alternatives need review. No selection, distinct-course audit, or new generic prerequisite sequence is implied.

A topological scheduler respects authored minimum semesters, strictly earlier and earlier-or-same edges, and the institution plan’s maximum load (16 at UNM). Upper-level nodes also respect upper_level_from_semester (5 in the pilot). Course groups have authored levels; concrete components must agree with the institution’s leading-digit bands. Never infer a group’s level from its display label. Earliest eligible planning slots win ties, then file order. This later upper-level stage is a user-requested planning preference, not catalog evidence. The full workload is the planning degree-credit target minus explicitly completed starting credits, plus preparation; this additional-work assumption is explicit, not an assertion that all preparation is ineligible for degree credit. Other work is the target minus shown non-preparation courses. The horizon grows to fit the workload and all scheduled courses. Other work fills toward the usual load (15), then uses the allowed maximum for the remainder. Semester totals include both named courses and other work. The latter is a workload allowance, within which shared rules earmark source-identified Gen Ed needs; other capacity stays elective/minor space.

build_degree_allowance_slots(terms, slot_credits) splits early allowances into course-sized grid tiles, using the plan’s allowance_slot_credits and preserving the exact budget. A usual slot can absorb one extra credit; smaller remainders remain explicit. It fills rows from the top of each semester. build_degree_roadmap() reserves the largest slot count as a shared top band and shifts all named course rows below it without disturbing math/writing alignment or interior course-family spans. A plan without early allowances reserves no extra rows. Later upper-level/ minor allowances remain below the grid. The plan returns the slots separately in allowances, never as catalog graph nodes; the view adds their display labels. start_season defaults to fall and alternates fall/spring labels, with a selector and shared-URL field. It affects example offerings, not prerequisite scheduling.

build_degree_popup_history() dispatches ordinary course cards to existing history and validates allowance clicks against the current plan. Suggestions use the captured annual area membership for identified Gen Ed cards, or lower/ upper undergraduate offerings for elective cards, excluding fixed/entry courses and finite option pools already represented in the program. get_gen_ed_planning_options() reads the explicitly supplied guarded history and section metadata; it never calls another cone. It selects non-topic, course-sized options from the supplied level scope and the latest matching-season row at each course/delivery-campus/college/session. Frequency denominators use source-covered terms only. Rows sort by frequency, then first-day enrollment; missing dates and suppressed counts stay unknown. Credit fit is an allowance comparison, not enrollment eligibility. The canonical format_degree_allowance_preview() prints the same aggregate payload for review. These are general historical examples, not a Gen Ed audit or curated advice for a major. Future major-specific Markdown guidance remains a separate extension. terms$other_phase distinguishes early foundation allowances from later advanced work. The feature labels identified area needs as Gen Ed and other early slots as elective/minor, any level. Late footer balances remain elective/minor context; explicit late Gen Ed cards are subtracted from those balances. No generic allowance is called an extra Gen Ed requirement.

Template positions and elective-group placements are planning suggestions. Course-specific options, offerings, admission, transfers, equivalences and the complete degree allocation are not evaluated. A missing edge does not assert that a course has no restrictions. Link/source validation rejects bad endpoints, cycles and unknown starting routes. The requirement tree and selected planning route are separate contracts; neither should silently update the other.

Degree planning uses metadata and redraws automatically. Its shared URL spec restores catalog_id, math_start and english_start; it has no run button.

Hover and keyboard focus preview named upstream/downstream chains and their course labels locally, without a history request. The floating preview is hoverable, dismisses with Escape and restores the last clicked chain on leave. Count conditions and unresolved alternatives retain their distinct semantics. Course clicks open the standard CEDAR modal (cedar_info_modal(scrollable = TRUE)), rendered by the server from degree_roadmap_tile_detail(view, node_id): catalog links at the top, enrollment history first and complete requirement evidence below. Escape or either close control returns focus to the tile without scrolling the page. Semester headers use identical fixed tracks.

get_course_planning_history(sections, students, subject_courses, edges) returns pressure at course/campus/college/part-term/term grain, plus terms, seasonal frequency, a complete-window offerings table, source coverage, missing exact codes and a window note. This is an all-course computation, not Regstats’ already-flagged subset. It reuses get_enrl(), filter_classlist_to_sections(), calc_first_day_enrl(..., by_part_term = TRUE) and enrollment_capacity_metrics(). Frequency uses all source-covered campus terms in the last nine possible terms through last_enrolled_complete. Missing source terms are excluded, never invented as non-offerings. Median seats sum sessions within an offered term; partial capacity leaves the total unknown.

Pressure uses reconstructed first-class-day enrollment and DESR scheduled capacity at matching campus/college/course/session/term grain. A term is near full if any session reaches 90%; seasonal pressure denominators require every session to be matched, plausible (fill at most 125%) and at least 10 first-day students. Missing matches are unknown, not zero enrollment; small counts and reconstructable percentages are suppressed. This describes potential planning pressure, not proven unmet demand, a frozen first-day roster or future course availability. Missing status, section-start or pull dates withhold the whole group’s first-day count. Banner’s current-status-only dates can overcount late additions that later dropped.

build_degree_course_history() resolves explicit course components and assembles one typed table. degree_history_table() uses the shared Reactable theme, cedar_count_coldef() / cedar_percent_coldef(), course/campus filters and cedar_pot_coldef(). Source-covered absence is Not offered (zero scheduled seats); missing coverage stays unknown. degree_course_history_ui(history, table_id) provides the table output slot and explanation; the module renders course_history_table separately. Unresolved course groups and new identifiers show explicit missing states. History is lazy and cached per clicked program/node in the Shiny session. Source tables, edges and catalog metadata are immutable session arguments, so placement changes need no history recomputation; a new session gets fresh data.

Catalog evidence and graph metrics

attach_degree_roadmap_evidence(nodes, roadmaps, catalog, catalog_id) resolves a tile’s actual reason for appearing. Degree membership must cite that program’s page; a course page may establish preparation but cannot substitute as degree proof. Every selected edge has a course prerequisite citation. Unknown IDs, missing tile membership, mismatched pages/types/years and malformed citations fail at startup. The source bar shows the exact saved excerpt and section link, plus separate program and course links. General English routes not explicitly named by Biology remain marked illustrative.

get_degree_chain_metrics(nodes, links, conditions = NULL) computes unique direct/indirect named ancestor counts, strict prerequisite depth, downstream reach and the longest sequence through a node. Before-or-concurrent edges carry zero additional earlier-term lag. Converging branches count ancestors once; paired lecture/lab nodes contribute two courses without adding a sequential term. Colors use ancestor counts (0, 1–2, 3+); unselected course groups with unknown prerequisites remain gray. Metrics concern the selected route after placement, not every catalog alternative or a student’s eligibility. Renderers consume them without recomputing graph rules.

Course-count and permission conditions live in their own file. The scheduler waits for enough eligible single-course slots to be scheduled, then places the target after the earliest qualifying completions. It never turns the whole pool into AND edges. The source must be the target course page. Conditions do not add named ancestors or named-chain depth; purple diamond labels display them and dotted outlines highlight qualifying options without arrows. Gen Ed suggestions derive from basis = general, with a visible label, light surface and dashed border.

One active catalog year is loaded. The text files are a reviewable annual graph snapshot, with catalog-year IDs and evidence; Git preserves previous sets. Live catalog hash URLs alone do not freeze an edition. The saved year, excerpt, retrieval date and review status record what was transcribed. Update the set as one reviewed annual change, retaining conflicts and checking degree/course pages independently. No live scraping or automatic inference runs in Shiny.

The offline catalog scanner captures complete source records, extracts draft trees and compares frozen editions. Its output remains separate from these reviewed files and authored semester templates. Collection and recognized syntax never confer academic approval or complete degree coverage.

Observed analysis

build_population_course_paths(students, population, groups, opt, description) uses the already defined Pathways population and calls get_grouped_student_paths(students, population, groups, opt). observedCoursePathsUI/Server() is mounted at Pathways > Course paths. The root passes edges and grouping metadata explicitly through pathwaysServer; its UI receives delivery-campus choices separately from home-campus choices. Changing the population invalidates the old result.

The cone enforces relevant_until, numbers full undergraduate participation history, then filters delivery campus and enrolled-term count. Registered and late-drop rows count. No grades or frozen cumulative credits are read. Course deduplication retains student, term, enrolled term, campus and course. Group cells and denominators retain campus and enrolled term; course detail also keeps course.

Numerator: distinct participants in any course of the group. Denominator: population students participating in any course at that delivery campus and enrolled term. Multiple groups can coexist, so shares need not sum to 100%. Band/cell guards default to 20/10. Suppressed counts become NA before the module sees them; known zeros stay zero only when the band is large enough. Course detail is filtered before returning. Unknown prefixes retain an unmapped group.

Term one is first observed study, not necessarily career entry. Left-boundary counts are disclosed. This is marginal course-mix timing, not individual paths or a graph of course-requisite or causal transitions. Subject categories are independent of credential requirements.

Forecast integration boundary

A future opportunity model should use reviewed rules to identify potentially needed coursework; observed histories estimate whether/when enrollment occurs. Do not sum alternatives as independent demand or use descriptive groups as eligibility rules. Resolve catalog version, transfer/waivers, equivalences, unresolved rules and overlap allocation before introducing a requirement-informed challenger in rolling aftcasts. Capacity remains an audit, never a predictor.

Regression tests live in tests/testthat/test-degree-explorer.R and run on a designed catalog, not on any institution’s real one: the demo institution’s institution/demo/degree_*.csv hold two synthetic programs (a calculus-route Biology B.S. and a statistics-route History B.A.) built to exercise every planning rule, and the file’s header works their schedules by hand. An annual catalog review therefore never breaks a unit test. One test still runs every starting scenario of the loaded institution’s catalog (UNM by default) and checks the planning invariants — credit ceiling, prerequisite timing, upper-level stage, no overlapping tiles — as an integrity check on the authored files. The standard degree-explorer browser suite covers placement changes, dependency shifts, paired courses, complete semester budgets, source evidence, planning colors, mobile scrolling, URL restoration and copying, and the moved observed view.

Shared requirement coverage in the semester board

load_degree_shared_requirements(catalog, institution_dir) loads the pinned annual bundle once in global.R as CEDAR_SHARED_REQUIREMENTS. The server passes this explicitly to Degree Explorer and Admin Visualization rules. degree_shared_coverage(view, roadmaps, reference) in branches consumes selected course components once across typed Gen Ed base/flexible areas; unresolved options advertise potential coverage without satisfying a requirement. Related science bundles remain intact. Completed-English input supplies only writing coverage; preparation eligibility and transfer/approval waivers are not inferred.

allocate_shared_degree_slots(view, coverage) earmarks named gaps within the existing per-term allowance budget, preserves remainders and exposes needs that cannot fit a whole slot. It shifts the common option band without altering named course timing or workload. footer_credits is the remaining late-term balance after any explicit late Gen Ed slots; rendering must not also count those slots in the footer. Other early cards are elective/minor space, any level.

The shared payload also shows the standard college upper-credit minimum, named upper-level credits and the balance still to fit in existing elective/minor space. Mixed clinical blocks are not counted wholesale. Conditional alternatives are recorded rather than selected. Typed planning interpretation fields live next to their source-linked annual rules; see shared_requirements/README.md. Flexible coverage is visibly provisional until the catalog numbering conflict is resolved. This is an illustrative allocation, not an eligibility check or degree audit.

The popup feature uses the selected area’s current captured memberships for a Gen Ed card, and all lower/upper undergraduate offerings for an elective card. get_gen_ed_planning_options(..., levels) retains campus/college/session grain and exposes course level. No enrollment metadata is treated as prerequisite evidence. Existing numeric Gen Ed analytics mappings remain separate.