Skip to content
imibare

Tutorial

Your first dataset

In this tutorial we will install imibare, load Rwanda’s monthly CPI data, and produce a simple chart in about five minutes. No prior knowledge of the project is required.

What we will build

By the end of this tutorial we will have a working Python script that:

  1. Loads Rwanda’s Consumer Price Index from 2009 to the present
  2. Prints a summary of the data
  3. Plots the overall CPI trend over time

Before we start

You need:

  • Python 3.11 or later
  • pip (comes with Python)
  • matplotlib (we will install it below)

Step 1: install imibare

Terminal window
pip install imibare matplotlib

You should see output ending with:

Successfully installed imibare-...

For Iceberg time travel (loading a specific past version of a dataset), install the optional extra: pip install imibare[iceberg]

Step 2: load the CPI dataset

Open a Python file (or an interactive session) and type:

import imibare as imi


df = imi.load("rw.nisr.cpi.monthly")
print(df.head())

Run it. You should see something like:

        date category   value
0 2009-01-01  overall    83.4
1 2009-01-01     food    92.1
2 2009-01-01  housing    79.3
3 2009-01-01 transport   77.8
4 2009-02-01  overall    84.1

imi.load() downloaded the Parquet file from Cloudflare R2 and cached it at ~/.imibare/cache/. The next call is instant.

Step 3: inspect the data

print(df.dtypes)
print(df.shape)
print(df["category"].unique())

You should see:

date        datetime64[ns]
category            object
value              float64
dtype: object


(3640, 3)


['overall' 'food' 'housing' 'transport' ...]

Notice that date is already a proper datetime column. imibare handles type conversion for you.

Step 4: plot the overall CPI trend

import matplotlib.pyplot as plt


overall = df[df["category"] == "overall"].sort_values("date")


plt.figure(figsize=(12, 4))
plt.plot(overall["date"], overall["value"], color="#c8602a", linewidth=1.5)
plt.title("Rwanda CPI, overall index (base year 2014=100)")
plt.xlabel("Date")
plt.ylabel("Index value")
plt.tight_layout()
plt.savefig("rw_cpi.png", dpi=150)
plt.show()

You should see a line chart saved to rw_cpi.png. Notice the base year switch in early 2015: the index resets to 100 because NISR changed the base year from 2009 to 2014.

Step 5: browse all available datasets

datasets = imi.catalog(country="RW")
for d in datasets:
    print(f"{d.id:45s}  {d.name}")

You should see:

rw.nisr.cpi.monthly                            Consumer Price Index (monthly)
rw.nisr.gdp.annual                             GDP by expenditure approach (annual)
rw.bnr.fx.daily                                Official exchange rates (daily)
rw.minecofin.budget-execution.quarterly        Budget execution by ministry (quarterly)
rw.rssb.contributions.annual                   RSSB contributions & member counts (annual)

What we built

We installed the package, loaded a live dataset, inspected its structure, and produced a chart in five steps. The data came directly from Rwanda’s National Institute of Statistics, processed and published by the imibare pipeline.

What’s next

  • Load a different dataset: try imi.load("rw.bnr.fx.daily") for daily exchange rates
  • Filter by date range: imi.load("rw.nisr.cpi.monthly", start="2020-01", end="2023-12")
  • Load as a Polars DataFrame: imi.load("rw.nisr.cpi.monthly", engine="polars")
  • Load a specific past version (requires imibare[iceberg]): imi.load("rw.nisr.cpi.monthly", version="2025-06-01")
  • See the full imi.load() API in the Python package reference
  • Query data directly with DuckDB: How to query with DuckDB

You can also skip the Python package entirely and query the same data straight from R2 with DuckDB’s Iceberg integration:

import duckdb


duckdb.sql("""
    ATTACH 'https://catalog.cloudflarestorage.com/imibare-data'
    AS imi (TYPE ICEBERG, READ_ONLY)
""")
duckdb.sql("SHOW TABLES").show()

See How to query with DuckDB for the full setup, including credentials.