Unit III — Data Structures & File Handling

Lists Hold Batches

Index map - five tablet weights (mg)

Positions: 0 1 2 3 4 left to right; from the right: -5 ... -1. w[0] is first, w[-1] is last. Off-by-one is the classic viva trap - index 5 does not exist here.

# Batch weights in mg - synthetic teaching data
weights = [480, 495, 505, 512, 498]
print(weights[0]) # first tablet
print(weights[-1]) # last tablet
print(len(weights)) # batch size

Slicing cuts a sub-batch in one move

The slice rule w[start:stop:step]

start is included, stop is excluded, step defaults to 1. So w[1:4] takes positions 1, 2, 3 - three items, never the 4th. w[::-1] reverses the whole batch.

# Same batch: [480, 495, 505, 512, 498]
print(weights[1:4]) # middle three
print(weights[:2]) # first two
print(weights[::-1]) # reversed QC check

Grow, shrink, sort and count the list

Four verbs that run a dispensary list

append(x) adds one item at the end +1 · remove(x) deletes by value -1 · sort() orders in place · len() counts. append vs extend: append adds one box, extend pours in many.

# Restock program - synthetic teaching data
stock = [480, 495, 505]
stock.append(507) ; print(stock)
stock.remove(495) ; print(stock)
print(sorted([505, 480, 512]))

Tuples Seal the Record

Why immutable? The batch-record argument

A batch record (id, drug, mg) must survive audits untouched. Immutability is a tamper seal: the QC stamp, assay value and expiry cannot be silently overwritten the way a list can. Use tuples for facts, lists for work-in-progress.

# Sealed batch record - synthetic teaching data
batch = ('B-1042', 'Paracetamol', 500, 25.0)
print(batch[1], len(batch))
batch[1] = 'Aspirin' # forbidden!

Dicts Name Every Field

# Patient record + mini formulary - synthetic teaching data
rec = {'name': 'Asha', 'age': 34, 'drug': 'Metformin', 'dose_mg': 500}
print(rec['drug'], rec['dose_mg'])
formulary = {'PCM': 500, 'AMX': 250}
print(formulary['PCM'])

get, update and keys - the safe toolkit

Three methods, one warning

get(k, default) reads without crashing on missing keys · update({...}) adds or overwrites fields · keys() / values() / items() list the contents. Direct rec[missing] raises KeyError - in a dispensary, always prefer get.

# rec from previous page
print(rec.get('drug'))
print(rec.get('allergy', 'NKDA'))
rec.update({'stock': 120})
print(list(rec.keys()))

Strings Clean Drug Names

print(' paracetamol 500MG '.strip().title())
print('amoxi,500,cap'.split(','))
print('-'.join(['Amoxicillin', '500mg']))

NumPy Arrays

import numpy as np
doses = np.array([250, 500, 750])
print(doses, doses.mean(), doses.sum())
print(np.zeros(3).tolist(), np.arange(0, 10, 2).tolist())

Creation verbs to memorise

np.array([...]) from a list · np.zeros(3) / np.ones(4) placeholders · np.arange(0,10,2) steps like range · .mean() / .sum() / .max() summarise instantly. All outputs above are real executed results.

One line beats a loop - dose for five weights

Vectorisation in one picture

Loop version repeats the multiply five times in Python. Vectorised version writes w * 10 once and NumPy multiplies all five in compiled C - same answer, a fraction of the time on large wards.

import numpy as np
w = np.array([40, 50, 60, 70, 80])
doses = w * 10 # mg-per-kg rate
print(doses.tolist(), doses.mean(), doses.sum())
loop = [x * 10 for x in [40, 50, 60, 70, 80]]
print(loop)

Writing patients.csv

import csv
rows = [['pid','name','drug','dose_mg','stock'],
    ['P01','Asha','Metformin',500,120],
    ['P02','Ravi','Amoxicillin',250,8],
    ['P03','Meena','Paracetamol',500,45],
    ['P04','Kiran','Insulin',100,5]]
with open('patients.csv', 'w', newline='') as f:
    csv.writer(f).writerows(rows)

Reading back with DictReader - real file I/O

DictReader names every column

csv.DictReader(f) turns each row into a dict keyed by the header - row["drug"] instead of row[2]. Note: every value arrives as text, so wrap stock in int() before comparing.

import csv
with open('patients.csv') as f:
    recs = list(csv.DictReader(f))
print(len(recs), [r['name'] for r in recs])
print([r['name'] for r in recs if int(r['stock']) < 10])

Columns are fields, rows are patients

How to read any healthcare table

A column is one field measured the same way for everyone (drug, dose_mg). A row is one patient visit. The header row names the columns; everything below is data. Shape of our file: 4 rows x 5 columns.

Import + manipulate in two verbs

import: DictReader loads the file into a list of dicts · manipulate: filter int(stock) < 10, sort by dose, count by drug. Master those two and every small pharma dataset obeys you.

Formulary lookup + low-stock report

What the demo proves

A formulary dict answers dose lookups in O(1); a list-of-dicts scan flags every stock below 10. Lookup + filter + report - the three moves of dispensary software in a few lines.

# formulary lookup + low-stock report - synthetic teaching data
form = {'Metformin': 500, 'Amoxicillin': 250, 'Insulin': 100}
print('Dose:', form.get('Insulin', 'unknown'))
low = [r['name'] for r in recs if int(r['stock']) < 10]
print('Reorder:', ', '.join(low))

IndexError vs KeyError - never confuse them

weights = [480, 495, 505, 512, 498]
try: weights[5]
except IndexError: print('Bad index: only 0-4 exist')
try: rec['allergy']
except KeyError: print('Use get(): NKDA assumed')

The chooser table - one glance, full marks

Chooser rule in one line

Changing pile? list · Sealed fact? tuple · Named fields? dict. Audit = tuple, queue = list, record = dict - write that sentence and the 5-marker is yours.

Three drills examiners repeat

Drill 1 - negative-index slicing drill

# tail drills - synthetic teaching data
w = [480, 495, 505, 512, 498]
print(w[-3:]) # last three
print(w[:-1]) # drop last
print(w[-4:-1]) # middle three

Drill 2 - dict .get() with default vs KeyError

# safe lookup - synthetic teaching data
rec = {'drug': 'Metformin', 'dose_mg': 500}
print(rec.get('drug')) # safe hit
print(rec.get('allergy', 'NKDA')) # safe miss
print(rec['allergy']) # crashes: KeyError

Drill 3 - batch totals: comprehension vs loop

# double the stock - synthetic teaching data
stocks = [120, 8, 45, 5]
loop = []
for s in stocks:
    loop.append(s * 2)
comp = [s * 2 for s in stocks]
print(loop, comp, loop == comp)

Trace first

Drill 1: -3: takes the tail - Drill 2: get never crashes - Drill 3: same list, one line beats four. Synthetic teaching data.

Exam Points

• Indexing: first is w[0], last is w[-1] - off-by-one costs the mark.
• Slicing: w[1:4] takes 1,2,3 - stop excluded, never the 4th.
• Tuple: immutable - batch record; edit raises TypeError.
• Dict: get(k, default) avoids KeyError; direct rec[k] crashes.
• Strings: methods return new strings - strip / split / join / upper; capture the result.
• NumPy: w * 10 vectorises five doses at once; .mean() / .sum() summarise.
• CSV: writerows saves, DictReader reads dicts; values arrive as text - int() first.
• Program: formulary lookup + stock < 10 filter + join reorder line - practise tracing it.

Five questions - answer before you turn the page

• Q1. weights has 5 items - why does weights[5] raise IndexError?
• Q2. What does weights[1:4] return, and why only three items?
• Q3. Batch record must survive audit - tuple or list, and why?
• Q4. rec['allergy'] is missing - which call avoids KeyError?
• Q5. How does w * 10 in NumPy differ from a five-round loop?

Answers (cover, then recall)

• A1. Valid indexes are 0-4; 5 is one past the end.
• A2. [495, 505, 512] - stop index 4 is excluded.
• A3. Tuple - immutable, tamper-sealed for audit.
• A4. rec.get('allergy', 'NKDA') - default, no crash.
• A5. Same values; vectorised runs once in C, loop repeats in Python.

Unit III in one line

Unit 3 teaches you how to store pharmacy data (list / tuple / dict), compute it (NumPy) and keep it (CSV) - so batch, dose and stock work runs itself.

Pinpoints - Must Remember

3 exam predictions

• Trace w[-3:] and w[1:4] on a 5-item batch.
• 2. Missing key: get vs rec[k] - which crashes?
• 3. Rewrite the dose vector w * 10 plus low-stock filter stock < 10 from memory.

Mnemonic - ZERO GET SEAL

Zero-based, tail with minus-one; GET averts KeyError, SEAL the tuple, vectorise the rest. Say ZERO-GET-SEAL once and all ten hooks follow.