Brantford data projectHow many play structures did Brantford install in 2017? Two, says the AI index
The City of Brantford's Play Equipment Areas file lists 221 pieces of park play equipment, with an ID, a park, a surface and an installation year. We wrote the program an assistant typically produces, then checked it. Everything below, code and counts, is ours, from the file downloaded on 23 October 2026.
The draft's idea: build a dictionary from FacilityID to row, then count the rows in it by Year_Ins. It is the standard way to remove duplicates, and assistants reach for it often. The problem is the key. Our count finds 47 rows with a blank FacilityID. A dictionary can hold only one entry for the blank key, so 46 rows vanish without an error.
The program, with the draft and the two checks side by side:# Brantford play equipment: a lookup keyed on FacilityID, and the row count that exposes it (our teaching example)
import csv
from collections import Counter
rows = list(csv.DictReader(open("hub_play_equipment.csv", encoding="utf-8-sig")))
# The draft an assistant often writes: index the rows by their ID, then count installs by year.
by_id = {r["FacilityID"].strip(): r for r in rows}
draft_years = Counter(r["Year_Ins"].strip() for r in by_id.values())
# The human checks: does the index still hold every row, and what did it throw away?
blank = [r for r in rows if not r["FacilityID"].strip()]
print("rows in file", len(rows), "| entries in the draft index", len(by_id), "| rows with a blank ID", len(blank))
checked_years = Counter(r["Year_Ins"].strip() for r in rows)
for year in ("2017", "2018", "1990"):
print(year, "draft", draft_years[year], "checked", checked_years[year])
print("blank-ID rows installed 2017 or 2018:", sum(r["Year_Ins"].strip() in ("2017", "2018") for r in blank))
Its output on the City's file:rows in file 221 | entries in the draft index 175 | rows with a blank ID 47
2017 draft 2 checked 22
2018 draft 1 checked 17
1990 draft 62 checked 66
blank-ID rows installed 2017 or 2018: 37
Installs by year: the AI-style index against the full file (our run, 23 October 2026)| Year_Ins | Draft index says | Full file says |
|---|
| 2017 | 2 | 22 |
| 2018 | 1 | 17 |
| 1990 | 62 | 66 |
Check 1: size of the index
Compare the number of entries in the dictionary with the number of rows read. 175 against 221 is the alarm. It takes one line of code and catches every key collision, whatever caused it.
Check 2: what was lost
List the rows the index dropped. 37 of the 47 blank-ID rows were installed in 2017 or 2018, so the lost data is not random: it is mostly the newest equipment, perhaps added before IDs were assigned. The file does not say why.
Check 3: suspicious peaks
Even the full count has a question in it: 66 rows give 1990 as the install year, far more than any other year. That may be a real build-out or a placeholder for unknown dates. We flag it rather than chart it as fact.
Codes nobody has defined
Subtype holds short codes such as 2bay on 61 rows, SandP and Seni on 25 each, and Status and Condition are bare numbers. The layer carries no code list, so an AI that confidently expands Seni to senior or reads Condition 1 as excellent is guessing. Some Location values on blank-ID rows are also cut short, such as SpringG. Our rule: describe what the file holds, decode nothing it does not explain.
The City's map service shows the layer's data last edited in August 2018, so this is an old snapshot rather than today's parks. Used under the Open Data License - Brantford, version 1.0, which allows use including for commercial purposes (licence). Contains information licensed under the Open Data License - Brantford. The City of Brantford has not endorsed this page or our classes.
