Source document. These are findings about the official syllabus documents, so they use those documents’ semesters and course numbers. The courses themselves are studied on their own, in any order.

On this page

1. Summary
2. Findings that affect your marks
3. Defects in the document
4. Design and sequencing issues
5. What the five Semester V courses do well
6. Verification notes for this repository

A check of the official syllabus documents, not a transcription of them. Every finding below was verified against docs/syllabus-extracted.md (Semesters I–II), docs/syllabus-extracted-sem3-4.md (Semesters III–IV) and docs/syllabus-extracted-sem5.md (Semester V) and docs/syllabus-extracted-sem6.md (Semester VI); page numbers refer to the source PDF of the semester in question.

Read this before you start studying. Three findings — D1, D2 and D13 — name topics that are examined but not listed in the syllabus units. If you study only the unit lists, you will walk into those questions unprepared. D13 sits below with the document defects because that is what caused it, but its consequence is the same as the other two.

---

Summary

ID	Finding	Severity
D1	Bayes' theorem examined but absent from the units	High — affects marks
D2	Database triggers examined but absent from the units	High — affects marks
D3	Text truncated in three places in the official PDF	Medium
D4	Lab question numbering broken in all three DBMS experiments	Low
D5	Course 2 Unit 4 title does not match its content	Low
D6	Course 3 Unit 4 carries roughly double a normal unit's load	Medium — affects planning
D7	Course 3 Unit 5 fuses two unrelated subjects	Medium — affects planning
D8	Statistics lab never uses Python, though Python is taught the same semester	Medium — affects skills
D9	Conditional formatting duplicated across Course 1 Units 4 and 5	Low
D10	Three-semester gap between statistics and its first application	Medium
D11	Course 2 activity list has an orphaned entry	Low
D12	Bibliographies damaged in all five Semester III–IV courses	Medium — you cannot order the books
D13	Course 10 Objective 4 is a fragment, and GridFS and transactions are examined but absent from the units	High — affects marks
D14	Course 10 Unit 1's "Installation & Setup" topic has lost text	Medium
D15	Course 8 numbers its units differently from every other course	Low
D16	Course 8's objectives are unnumbered, unlike every other course's	Low
D17	Course 11's Objective 3 stops mid-sentence	Medium
D18	"Business IntelligenceI" — a stray capital in Course 11 Unit 1	Low
D19	Course 12 A has 3 objectives against 4 outcomes against 5 units	Medium — affects planning
D20	Course 11 numbers its units in Roman numerals	Low
D21	Course 12 B's reference books are numbered 3 and 4	Low
D22	Course 13 B's reference list has an empty entry 4	Low
D23	"SudoStudents" — two words merged in Course 13 A's activities	Low
D24	"HadoopSpark" and "HBaseSpark" — separators lost, twice	Low

D1–D11 come from the Semester I–II document, D12–D16 from the Semester III–IV document, D17–D24 from the Semester V document and D25–D33 from the Semester VI document. Every one was found by the same method — reading the extracted text rather than the formatted PDF, where truncation is invisible.

---

Findings that affect your marks

D1 — Bayes' theorem is examined but never listed as a topic

Course 4, Unit 1 lists the probability topics as: "Concept of Uncertainty, Axioms and rules of probability, Conditional probability". Bayes' theorem does not appear in Unit 1, or in any other unit of the course.

But the prescribed activities say: "Classroom Quiz (MCQs & short problems on probability, conditional probability, Bayes)". And lab experiment 2 is a fragment reading only "a positive result." — the surviving tail of the standard Bayes medical-testing problem ("…given that the test returns a positive result").

Consequence: a topic that carries assessment weight sits outside the syllabus list you would revise from.

What to do: study Bayes' theorem as though it were listed in Unit 1. It is covered in Course 4 Unit 1, where it is flagged as an off-syllabus-but-examined addition, with the reconstructed medical-test problem worked in full.

D2 — Database triggers are examined but never listed as a topic

Course 5, Unit 5 (PL/SQL) lists: "Introduction, Shortcomings of SQL, Structure of PL/SQL, PL/SQL Language Elements, Data Types, Operators Precedence, Control Structures, Steps to Create a PL/SQL Program, Iterative Control, Procedures, Functions." Triggers are not among them.

Yet triggers are required in three separate places:

• Course Objective 5: "…incorporating control structures, functions, procedures, and database triggers."

• The prescribed activities: "Build a PL/SQL-based payroll or student grading system using: Procedures and functions, Control structures (IF, LOOP), Triggers for automated updates."

• Lab Section E, items 5 and 6: "Create a trigger to prevent inserting employees with a salary less than 30,000" and "Create a trigger to avoid any transactions (insert, update, delete) on the EMP table on Saturday & Sunday."

Note also that Course Outcome 5 drops triggers even though the Objective names them — so the omission is consistent between the outcome and the unit, and inconsistent with the assessment.

What to do: study triggers as part of Unit 5. Covered in Course 5 Unit 5, with both lab triggers written out.

---

Defects in the document

D3 — Truncated text in three places

1. Course 4, Outcome 1 reads: "Apply the basic rules of probability, conditisolve problems involving uncertainty." Words have been dropped mid- sentence. The intended reading is almost certainly "…probability, conditional probability and Bayes' theorem to solve problems involving uncertainty" — which independently supports D1.

2. Course 4, lab experiment 2 is just "a positive result." The question stem is gone. Reconstructed in the Course 4 lab notes.

3. The elective note reads: "students are required to select a pair of electives from one of the TWO specified domains. is chosen, courses 12 to 15 to be chosen as 12 A, 13 A, 14 A and 15 A." The clause naming the two domains has been lost, so the domains are never actually named anywhere in the document. The A/B tracks can be inferred from the course titles (see SYLLABUS-MAP.md §1), but confirm the official names with your department before committing to a track in Year III.

D4 — Lab question numbering is broken in all three DBMS experiments

Verified by extracting every numbered item:

Experiment	Pages	Missing numbers
1 — Inventory Management	29	3, 13, 20, 22
2 — Online Bookstore	31–32	12, 19
3 — Employee DB	36–37	8
3 — Section E (PL/SQL)	37	2

Some gaps lost their text entirely; others left orphans. PL/SQL item 2 is the clearest case — its text survives as a dangling fragment, "If yes, print 'High Salary'; Otherwise print 'Standard Salary'", with no question in front of it. The intended question is evidently "write a PL/SQL block that checks whether an employee's salary exceeds some threshold".

Section B of Experiment 1 also begins at 4, and items 6 and 8 are cut short ("Update the stock quantity of" — of what?).

What to do: the lab notes renumber cleanly and reconstruct the missing items, marking each reconstruction as such so you can tell it from the official text.

D5 — Course 2 Unit 4 is titled "Functions" but opens with pointers

The unit is headed "Unit 4. Functions:" and its first topics are "Pointers: Pointer data type, Pointer declaration, initialization, accessing values using pointers. Pointer arithmetic, Pointers and arrays." Functions follow after pointers, then storage classes.

Harmless to the content, but if you revise from unit titles you will not expect pointers to be examined under "Functions". The notes cover them in the order printed and flag the mismatch.

D11 — Orphaned activity in Course 2

Every activity in the document follows the pattern Outcome: … / Activity: … / Evaluation Method: …. The "Recursive Problem Solver" activity appears with no Outcome: heading above it, unlike its four siblings — its outcome line was dropped. It maps to Course Outcome 4 (modular code using functions, recursion and parameter passing).

D12 — Bibliographies are damaged in all five Semester III–IV courses

Every one of the five courses has at least one book entry that is unusable as written. This is worse than it sounds: a bibliography exists so you can order the book, and none of these five entries names one you could order.

Course	Page	What the document says	What is missing
6 — Data Science with R	5	Textbooks 1. …Springer, 2nd Edition, 2021 2. Reference Books	Textbook 2 is empty — the number survives, the entry does not
7 — Web Technologies	9	4. JSON at Work Media. Reference Books 1. 2. An Introduction to HTML…	Textbook 4 has lost its author (Tom Marrs) and its publisher is a bare "Media"; reference book 1 is empty
9 — Python for Data Analysis	19	1. …pandas, NumPy, and Jupyter, Wes 2. Python Programming…	The author is cut mid-name at "Wes" — Wes McKinney
10 — Document Oriented Database	23	Textbooks: 1. MongoDB: The Definitive Guide, …Chodorow, 2. MongoDB Recipes…	Textbook 1 has no publisher, edition or year
10 — Document Oriented Database	23	Reference Books: 1. MongoDB in Action… 2. 3. M Web Resources:	Reference books 2 and 3 are gone, leaving an orphaned "M"

What to do: the four books that are fully named are the ones to buy. Each course README in this repository lists them with the missing details filled in and marked as reconstructed, so you can tell my additions from the document's own text.

D13 — Course 10's Objective 4 is a fragment, and GridFS and transactions go missing with it

The five course objectives read:

Objective 4 has lost everything before "replication" — no verb, no capital, no sentence. Course Outcome 4 on the same page survives intact and shows what was intended: "Utilize advanced features like indexing, aggregation, GridFS, and transactions to optimize data handling." So the objective was almost certainly "To familiarise students with indexing, aggregation, GridFS, replication, and transactions."

Topic	Appears in	Appears in a unit?
GridFS	Outcome 4; lab experiment 18	No
Transactions	Outcome 4; the truncated Objective 4; lab experiment 19	No

So a student who revises from the unit lists alone meets neither, and then finds two of the twenty lab experiments are about them. That is the same shape of problem as D1 and D2. Both are covered in Unit 5 and in lab experiments 18 and 19 for that reason — and both of those lab files also say plainly that they need a running server, which is the other half of the answer when an examiner asks you to demonstrate them.

D14 — Course 10 Unit 1's "Installation & Setup" topic has lost text

Unit 1's topic list ends across the page break as:

Two words have been welded together ("Setup" + "service"), and a closing parenthesis survives with no opening one — so an entire parenthetical clause has been dropped. Judging by the surviving fragment it listed the install routes and named the server process, something like "Installation & Setup (installing MongoDB, starting the mongod service), connecting via Mongo shell or GUI."

Unit 1 §1.9 covers what it evidently intended — Atlas, Docker and a local install, the mongod service, and connecting with mongosh or Compass — and says in the section itself that the source text is damaged.

D15 — Course 8 numbers its units differently from every other course

Data Mining heads its units Unit-1:, Unit-2: … with a hyphen and a colon. All nine other courses in both documents use Unit 1., Unit 2. … with a space and a full stop.

Cosmetic, and listed only because it is the kind of inconsistency that suggests this course's text was pasted in from a different source document — which is also the most likely explanation for D16.

D16 — Course 8's objectives are unnumbered, unlike every other course's

Every course in both documents numbers its five objectives 1–5. Course 8 runs them together as unnumbered sentences:

There are still exactly five, and they still map one-to-one onto the five outcomes and the five units, so nothing is missing — but if an examiner asks you to "state the third course objective", the document does not tell you which one that is. Counting in printed order gives association rule mining, which is also Unit 3 and Outcome 3.

---

Design and sequencing issues

D6 — Course 3 Unit 4 carries roughly double the load of a normal unit

Unit 4 is "File Handling, Exception Handling & Object Oriented Programming" and contains:

• file types, paths, open/close, read/write, CSV, os/pathlib

• syntax errors, built-in exceptions, try-except, raise, user-defined exceptions, assertions

• classes, objects, attributes, methods, constructors, destructors

• encapsulation with private and public members
• inheritance — single, multilevel and multiple — plus method overriding

That is three teachable units compressed into one. OOP alone is normally a full unit. Compare Unit 1, which covers only literals, variables and operators.

What to do: budget roughly twice the study time for Unit 4 as for Unit 1. The study plan already does this, splitting it across three weeks.

D7 — Course 3 Unit 5 fuses two unrelated subjects

Unit 5 is "Abstract Data Structures and GUI Programming" — linked lists, stacks, queues and priority queues, and Tkinter widgets and event handling. These share nothing conceptually. Data structures are algorithmic and carry the exam weight; Tkinter is applied and carries the lab weight (2 of the 18 lab programs).

What to do: treat them as two separate topics. The notes split unit-5.md into two clearly divided halves.

D8 — The statistics lab never touches Python

Course 4's lab is headed "Advanced Spreadsheets/Excel Lab/PSPP Open Source", and all 15 experiments are spreadsheet exercises — NORM.DIST, NORM.INV, EXPON.DIST, the Data Analysis ToolPak, the Regression tool.

Meanwhile Course 3 teaches Python in the same semester. The two courses never meet. Python-based data analysis waits until Semester IV, Course 9 ("Python for Data Analysis and Visualization").

This is defensible pedagogically — a spreadsheet makes the arithmetic of variance or a t-test visible in a way scipy.stats.ttest_ind() does not. But it means you finish Semester II able to compute a regression in Excel and not in the language you just spent a semester learning.

What to do: do each experiment twice — once in Excel exactly as prescribed (that is what the exam tests) and once in Python (that is what the degree is for). The Course 4 lab notes give both versions of all 15.

D9 — Conditional formatting appears twice in Course 1

Unit 4 lists "Data Handling: Sorting, filtering, conditional formatting". Unit 5 then opens with "Conditional Formatting: Custom rules, Color scales, Icon sets, Data bars" as a headline topic. Minor, but worth knowing that Unit 5's treatment is the substantive one.

D10 — Three-semester gap between statistics and its first real application

Regression, correlation and hypothesis testing are taught in Semester II (Course 4). Their first genuine application is Data Mining in Semester IV, and Machine Learning is a Year III elective in Semester V — three semesters after the theory.

Statistical intuition decays without use. The study plan schedules a short Course 4 refresher before Semester IV begins, so Data Mining does not start with re-learning what a p-value is.

D17 — Course 11's Objective 3 stops mid-sentence

The third objective of Business Intelligence Tools reads:

and then stops. The verb, the object and the full stop are all missing.

What was almost certainly there: "…using Power Query and DAX", since DAX is the whole of Unit 3 and Outcome 3 names it. But that is an inference, and if an examiner asks you to state Objective 3 the document cannot tell you.

This is the third truncation of the same kind — see D3 (three in the Semesters I–II document) and D13 (Course 10's Objective 4). Four truncations across two documents in the same position — the end of a numbered objective — suggests a systematic problem in whatever produced the PDF rather than four independent typing errors.

D18 — "Business IntelligenceI" — a stray capital in Course 11 Unit 1

Unit 1's topic list reads:

A stray capital I is welded onto "Intelligence" the second time it appears. Trivially cosmetic — but it sits in the first line of the first unit of the first course of the semester, which is the last place a proofreading pass should miss.

D19 — Course 12 A has three objectives against four outcomes against five units

Every other course in both documents has five objectives and five outcomes, mapping one-to-one onto five units. Machine Learning has three objectives and four outcomes — and five units.

Course	Objectives	Outcomes	Units
11	5	5	5
12 A	3	4	5
12 B	5	5	5
13 A	5	5	5
13 B	5	5	5

All four outcomes do have activities and evaluation methods, so nothing is left outside continuous assessment. What breaks is the mapping: Outcome 4 carries both Unit 5's clustering algorithms and "identify suitable machine learning approaches for specific application domains", which is the case-study material. Two units' worth of content, one outcome.

The practical consequence for a student: Unit 5 is the largest unit in the course by topic count — k-Means, k-Medoids, hierarchical clustering, DBSCAN, internal and external validation, and four case studies — and it is weighted in the outcomes as half of one. Do not read that as permission to give it half the time. The Course 12 A notes give it the same weight as the other four.

And Outcome 4 has no full stop — it runs straight into "Unit 1." on the same line — which is the same boundary defect as D17 and D23, in a milder form.

D20 — Course 11 numbers its units in Roman numerals

Business Intelligence Tools heads its units Unit-I:, Unit-II: … Every other course in the Semester V document uses Unit 1. with an Arabic numeral, a space and a full stop.

This is D15 again in a different course — Data Mining used Unit-1: where everything else used Unit 1. — and it is the same signal: these course descriptions were assembled from separately-authored sources and never passed through a single formatting pass.

D21 — Course 12 B's reference books are numbered 3 and 4

The bibliography reads:

The reference list continues the textbook numbering instead of restarting at

1. Harmless in itself — but D12 found the Semester III–IV bibliographies damaged in all five courses, and this is the same section behaving oddly again.

D22 — Course 13 B's reference list has an empty entry 4

The list runs:

Entry 4 is a number with nothing after it. A book was removed, or failed to paste, and the numbering was never closed up. The list claims five references and supplies four.

D23 — "e.g., SudoStudents" — a merged word in Course 13 A's activities

The Outcome 4 activity reads:

"Sudoku)" and "Students" have merged into "SudoStudents", taking the closing bracket with them. The intended text is almost certainly "(e.g., Sudoku). Students write…".

Same defect class as D3, D13, D14 and D17, and the fourth instance of text being lost at a boundary in this pair of documents.

D24 — "HadoopSpark" appears twice in Course 12 B

Objective 5 and Outcome 5 both read "HadoopSpark integration", and the case-study activity reads "HBaseSpark integration".

The missing character is presumably a hyphen or a slash — "Hadoop-Spark", "Hadoop/Spark". Two words joined where a separator was lost, which is the same mechanism as D23 and D14.

Counting the whole class: across the two documents there are now nine places where text has been lost at a word or sentence boundary — D3 (three), D13, D14, D17, D23, and D24 (two). That is no longer a scattering of typos; it is a defect in the production of the documents, and the honest summary is that both PDFs need a proofreading pass before they are issued to students.

---

D25 — "design an end-to-ple:" — a truncated sentence in Course 14 B

The activity for Outcome 4 reads:

The intended text is almost certainly "design an end-to-end pipeline", and "pipeline" has been chopped to "ple:" — the colon is what survived of the word plus the punctuation that followed it. This is the most severe of the truncation class because the sentence no longer parses at all.

D26 — "Coders(Based" — a missing space in Course 14 A's references

Reference 2 reads:

A space has been lost before the opening parenthesis. Same mechanism as D23 and D24 — a word boundary that did not survive whatever produced the PDF.

D27 — Course 15 A numbers its second reference "2. 2."

The reference list reads:

A duplicated list marker. Cosmetic, but it is the third distinct list-numbering defect in the programme, after D21 (Roman numerals in Course

11) and D22 (an empty entry 4 in Course 13 B).

D28 — "Tensor Flow" and "Mobile Net" split in Course 14 A's practicals

Practical 2 reads "Google Teachable Machine or Tensor Flow Playground" and practical 8 reads "Fine-tune a pre-trained CNN (Mobile Net, VGG)".

These are the inverse of D23, D24 and D26: a space inserted into a proper noun rather than lost between two words. Both product names are single words — TensorFlow, MobileNet. That the same document contains both failure modes points at automated text extraction rather than typing.

D29 — A "short reflective note (12 pages)" in Course 14 A

The evaluation method for the Unit 5 activity asks for:

A "short" note of twelve pages is a contradiction, and it is almost certainly "1-2 pages" with the hyphen lost — the same class of defect as D25, D26 and the rest, but here it changes the workload a student is being asked for by a factor of six to twelve. This one has a practical consequence, not just a cosmetic one: a student reading it literally would write twelve pages, and a marker expecting one page would not know why.

D30 — Course 14 B's evaluation names R and Python; the practicals assume both

The course objectives say "hands-on practice with real-world datasets using R/Python statistical libraries", and the practical list is headed "Using R/Python statsmodels, pandas, forecast, or equivalent".

statsmodels and pandas are Python; forecast is an R package. The practical list mixes the two ecosystems in a single line without saying which a student should use, and the department has not stated which will be available in the lab.

This is not a typo but a specification gap, and it matters: the two languages differ in how they handle seasonal periods, in what auto.arima does that statsmodels does not, and in the default confidence-interval method. This repository resolves it by using Python throughout, and says so — but a student following the syllabus literally has no basis for the choice.

Updating the count from D24: across the four documents there are now fourteen places where text has been lost or inserted at a word or sentence boundary, spread over ten findings — D3 (three), D13, D14, D17, D23, D24 (two), D25, D26, D28 (two), and D29. Semester VI contributes five of them in seventeen pages — one defect per 3.4 pages, and it is the shortest of the four documents.

The recommendation from D24 stands and strengthens: all four PDFs need a proofreading pass before they are issued to students.

D31 — Course 15 B's only textbook has no publisher and no year

The reading list reads, in full:

The trailing comma after "Housley" is where the publisher and year should be — the book is O'Reilly, 2022 — and item 2 turns out not to be a book at all but the heading "Web Resources", so the list has exactly one text and it is incompletely cited.

Compare this with the other Semester VI courses, which give author, year, title, edition and publisher. Course 15 B is the only one whose students cannot locate their single prescribed text from the syllabus alone.

D32 — Course 15 B prescribes a paywalled article as a course resource

Among the three web resources is a Towards Data Science article. That publication moved to a Medium members-only model, so a student following the syllabus may find the prescribed reading behind a paywall.

This is a different class of defect from the truncations. It is not a production error — somebody chose that link — but it is a maintenance failure: a syllabus that cites a URL has taken on the obligation to check that the URL still serves what it did. The IBM and Martin Fowler links alongside it remain freely readable.

D33 — The Semester VI practical lists name tools the theory never mentions

Three cases, across both tracks:

Course	The practical asks for	The theory units mention
14 A, practical 2	Google Teachable Machine, TensorFlow Playground	neither
15 B, practical 4	Apache Kafka or RabbitMQ	"event-driven architecture", no product named
15 B, practical 15	Prometheus / Grafana	"logging, monitoring frameworks", no product named

Practical 4 is the significant one. A student is asked to compare batch and event-driven ingestion using Kafka or RabbitMQ having been taught the architectural concept and nothing about either product — not the broker model, not partitions, not consumer groups, not delivery semantics.

The lab is being asked to teach the tool as well as apply it, in two hours, and the syllabus allots no theory time to it. Either the unit should name the technology it expects, or the practical should say that learning it is part of the exercise.

---

What the five Semester V courses do well

The defects above are worth recording, but a review that lists only faults misrepresents the document. Four things in Semester V are genuinely well done.

The elective tracks are coherent. Track A (Machine Learning → AI → Neural Networks → NLP) and Track B (Big Data → Cloud → Time Series → Data Engineering) each build properly, and the constraint that you stay in one domain across both semesters is correct — the Semester VI courses genuinely depend on their Semester V predecessors.

Course 12 A's unit ordering is right. Putting model preparation and evaluation (Unit 2) before any algorithm is unusual and correct. Most syllabi teach algorithms first and evaluation last, which is how students learn to quote accuracy without a base rate.

Course 13 B ends on monitoring and cost. Unit 5 covers deployment, monitoring, scalability and cost optimisation as examinable content rather than as an afterthought. That is more realistic than most cloud syllabi, which stop at "deploy the model".

Course 12 B pairs every high-level tool with its foundation. MapReduce before Hive, HDFS before HBase, and the Spark comparison stated in terms of what MapReduce does between stages. A student who follows the order will understand why Spark replaced MapReduce rather than merely that it did.

---

Verification notes for this repository

The lab code in labs/ was checked as follows. Run bash tools/verify_all.sh to reproduce all of it.

What runs, and is asserted

Course	Language	Status
2	C (15 programs)	Compiled and run. gcc -Wall -Wextra, no warnings, output verified against expected results.
3	Python (all 18)	Run. Python 3.11; the two Tkinter programs under a virtual display, their buttons pressed by a driver.
4	Python (15 equivalents)	Run, and their results checked against the notes' hand-computed figures.
5	SQL — DDL/DML/queries	Executed against SQLite via tools/run_sql_labs.py, with schema and the official sample data loaded.
6	R (18 scripts) and Python (14 equivalents)	Run. R 4.3.3, via tools/run_r_equivalents.py; the plotly charts and the Shiny app are opened in Chromium. Python: one per R script that has a counterpart.
7	JavaScript + DOM (16 experiments)	Run under jsdom, 184 assertions on the resulting DOM state, via tools/run_web_labs.js.
8	WEKA 3.8.7 (14 scripts) and scikit-learn / mlxtend (15 experiments)	Run, and every hand trace in the notes — Apriori's itemsets, ID3's information gains, K-Means to convergence — reproduced by executing code.
9	NumPy / Pandas (18 practicals)	All run, outputs asserted. Nothing in this course is desk-checked.
10	MongoDB 8.3.7 (all 20 scripts) and mongomock (16 of 20)	Run, every query and pipeline asserted, via tools/run_mongo_labs.py; replication on a real replica set, GridFS with mongofiles.
11	Python (BI semantics)	Run via tools/run_bi_labs.py. Every DAX, Power Query and LOD figure in the notes was computed, not quoted.
12 A	scikit-learn (12 practicals)	All run via tools/run_ml_labs.py. No file in this course is marked NOT EXECUTED — nothing it needs is blocked.
12 B	Hadoop 3.3.6, Pig, Hive, Sqoop, Flume, HBase, ZooKeeper (15 tool files); Python, DuckDB, fastavro, pyarrow, PySpark (14 programs)	Run via tools/run_bigdata_labs.py: every tool file on a real cluster, and a real SparkSession with a real shuffle, and real Avro and Parquet files.
13 A	SWI-Prolog 9.0.4 (16 .pl files); pytholog + Python (7 programs)	Run via tools/run_ai_labs.py: every .pl file's queries in SWI-Prolog, and five experiments as real logic programs through pytholog's SLD resolution.
13 B	Python, DuckDB, scikit-learn, http.server (7 programs, 15 experiments)	Run via tools/run_cloud_labs.py, including a real web server, a real ETL into a real columnar warehouse, and a real REST endpoint serving a real model.

What does not run, and says so in its own file header

Updated October 2026: R and tkinter are now installed here, so the R scripts and the two Tkinter programs, listed below until then, now run. So, later that month, do WEKA (Course 8), MongoDB (Course 10), the Hadoop stack (Course 12 B) and SWI-Prolog (Course 13 A), whose rows are removed.

Course	What	Why, and what stands in for it
4	Excel and PSPP walkthroughs	Not executable. Written as step-by-step instructions with exact formulas; the Python equivalents of the same 15 experiments were run.
5	PL/SQL (procedures, functions, triggers)	The syllabus targets Oracle PL/SQL; SQLite cannot run it and no Oracle instance is available. Written to Oracle syntax and reviewed by hand — verify on your college's installation.
11	Power BI and Tableau (all 15 click-paths)	Neither tool runs on Linux, and Tableau Public needs an account. Each experiment documents the exact click-path; the semantics behind every figure were computed in Python.
13 B	AWS, Azure, GCP and VMware (14 files)	No cloud account exists for this repository and none was created — signing up requires a payment card and accepts a billing relationship. Every provider claim is documented, never demonstrated.

The rule the whole repository follows: every numeric claim in the notes is checked by running code, and anything that genuinely cannot run says NOT EXECUTED in its own first lines rather than implying a test that never happened.

That discipline found more than fifty errors of my own while these notes were being written — four in Courses 1–5, three in Course 6, six in Course 8, seven in Course 9, two in Course 10, and roughly thirty more across the five Semester V courses. Every one would otherwise have been a wrong worked example a student revised from.

And it caught more than arithmetic. Several results came out contradicting what I had expected to write, and the notes report the measurement rather than the expectation:

• Random forest lost to a single decision tree on an iris split (0.8889 against 0.9778), and the cross-validated figures overlap almost entirely — so the honest conclusion is that the models are not distinguishable on that data.

• Scaling made k-NN worse on iris — 0.9778 unscaled against 0.9111 scaled — even though it transformed a synthetic case from 0.5500 to 0.9750. Iris's four features already share a scale, so standardising only discarded information.

• Adjusted R² rose by chance in 119 of 300 trials, so a single-dataset demonstration of "adjusted R² penalises useless predictors" would have been luck rather than evidence; the notes run the distribution instead.

• Hive's three-bucket hash left one bucket empty with three store names.

• The MapReduce combiner saved only 18.75%, because the splits are tiny — which is the honest way to teach that its value scales with split size.

• Autoscaling cost more than fixed capacity in one measured configuration (188 instance-hours against 168), directly contradicting the slogan.

• pytholog gives a wrong answer for cousin/2 through a nested derived predicate, returning a person as his own cousin — documented as an engine limitation, with the correct flat formulation shown alongside.

• Parquet is 4.8× LARGER than CSV at nine rows, and its 303× advantage on repeated data falls to 12× once every row differs.

Reporting a result that undercuts the point you were about to make is the whole value of executing the code. A note that only ever confirms its own claims has not been checked; it has been decorated.

tools/check_coverage.py is held to the same standard, and it had to be fixed before it could be trusted: it was matching its keywords against the syllabus line each note file quotes in its header, so 19 of 464 topics were passing without a word being written about them. Stripping the header before searching exposed three genuine gaps in Courses 1, 3 and 7, and three more in Course 10. All six are now written. Two further fixes followed as Semester VI was added: the checker now folds -ise/-ize spellings and en-dashes, and collapses whitespace, because a keyword that happened to straddle a line wrap was being reported as a missing topic. The check currently reports 1,273 of 1,273 topics across 95 unit files.

And the harness itself failed, in the way harnesses do

tools/verify_all.sh reported ALL VERIFICATIONS PASSED while six lab programs were failing underneath it.

Every suite in that script is piped into tail so the summary stays readable:

python3 "$ROOT/tools/run_data_labs.py" | tail -3 || fail=$((fail+1))

A pipeline's exit status is the status of its last command, which is tail, which always succeeds. The || branch could never fire. The fix is one line — set -o pipefail — and the script had run green for the whole project without it.

What it was hiding was real: installing mlflow and dvc for Course 15 B pulled pandas back from 3.0.5 to 2.3.3, leaving two conflicting dist-info directories in the same location. Courses 8, 9 and 12 A are written against pandas 3 — they assert that groupby().apply() drops the grouping column and that chained assignment now fails silently rather than warning — so five programs in Courses 8 and 9 and one in Course 12 A began failing. Restoring pandas 3.0.5 returned both suites to 33 of 33 and 12 of 12, and tools/requirements.txt now pins pandas>=3.0 with the reason recorded.