Appendix C — Glossary

This appendix gathers 335 technical terms from the whole book, in alphabetical order. Each term has a simple meaning and the number of the chapter where it is explained. When a term is explained or introduced in two chapters, both are given (for example “Chapters 28, 60”); the later one is usually the deeper treatment.

How to use the glossary

  • Look up the first letter of the term: A B C D E F G H I J L M N O P Q R S T U V W X Y Z.
  • Terms that start with punctuation (such as .env) are sorted by their first letter after the punctuation (env, under E).
  • The meanings here are deliberately short and simple. The full explanation, code examples, and their output are in the chapter referred to.
  • A term missing here may be explained directly in the relevant chapter; open the table of contents, or use search in the web version.
NoteREMEMBER

The chapter numbers in this glossary were checked against the manuscript with a checker script: for each term, the script looks for that word in the chapter named, in both the Indonesian and the English versions. If a term is not found in that chapter, this list is not built. The script confirms that the term appears in the chapter referred to; which chapter counts as “where it is explained” remains the author’s judgement.

List of terms

A

AI agent
A system that uses a language model to observe, plan, act, and evaluate with tools. Chapter 62.
algorithm
A clear, unambiguous sequence of steps for solving a problem. Chapter 2.
allometric equation
A formula estimating tree biomass or volume from easily measured sizes such as diameter. Chapter 59.
Anaconda
A Python distribution with many data-science libraries and the conda manager. Chapters 4, 17.
ANOVA
A test for comparing the averages of more than two groups. Chapter 26.
API
The official door that lets one program ask another for data or services. Chapter 21.
API key
A secret code that marks who is using an API. Chapter 21.
argument
The value you give when you call a function. Chapter 10.
array
A collection of numbers of one type in numpy, like a sack of identical items. Chapter 23.
assert
A statement that stops the program if a condition you believed true turns out false. Chapter 12.
AST (abstract syntax tree)
The abstract syntax tree: a structured form of source code that can be inspected. Chapter 64.
asyncio
The module for concurrency with async and await within one thread. Chapter 35.
attention
A mechanism where each word weighs the other words to understand its context. Chapter 61.
augmentation
Adding variety to training data by rotating, flipping, or changing image brightness. Chapter 56.
autograd
The PyTorch feature that computes gradients automatically. Chapter 55.
automated test
Small code that checks other code works right, run with a single command. Chapter 18.

B

band
One layer of values in a raster, for example red or near infrared. Chapter 39.
biomass
The mass of living material, for example a tree’s above-ground weight. Chapters 6, 59.
boolean
A type that is only True or False. Chapter 5.
boolean mask
An array of true/false used to filter rows. Chapter 23.
branch
A separate line of history for working on changes without disturbing the main line. Chapter 20.
break and continue
break stops the loop; continue jumps to the next round. Chapter 8.
breakpoint
A point where the program pauses so it can be inspected step by step. Chapter 19.
broadcasting
The numpy rule for combining arrays of different shapes in one operation. Chapter 25.
broker
An intermediary that stores and forwards messages or tasks, for example Redis. Chapters 52, 67.
buffer
A zone at a given distance around a shape. Chapter 38.
bug
A flaw in a program that makes its result differ from what was expected. Chapter 12.
bytecode
The intermediate instructions compiled from Python code that the virtual machine runs. Chapter 65.

C

caching
Keeping a computed result so it need not be computed again. Chapter 44.
callback
A function called automatically when an event happens, for example a click. Chapter 43.
Celery
A popular task queue library in Python. Chapter 67.
checksum
A fingerprint of a file to confirm its content has not changed. Chapter 29.
chi-square
A test of the relationship between two categorical variables from a count table. Chapter 26.
chunk
A small piece of data processed one at a time or in parallel. Chapter 53.
CI/CD
An automatic workflow that tests code (CI) and releases it (CD) on every change. Chapter 48.
class
A mold for making objects. Chapter 16.
CLI (command-line interface)
A program used by typing commands and options in the terminal. Chapter 42.
closure
A function inside a function that remembers variables from its wrapper. Chapter 31.
cloud-native
A data format that can be read in part directly over the internet without downloading everything. Chapter 51.
CNN
A convolutional neural network for images. Chapter 55.
COG (Cloud Optimized GeoTIFF)
A GeoTIFF arranged so it can be read in small windows over the internet. Chapter 51.
comment
A note for humans inside the code, starting with #, which Python ignores. Chapter 4.
commit
One record of changes saved permanently in Git’s history. Chapter 20.
comparison operator
An operator that compares two values and gives True or False, like == or >. Chapter 6.
comprehension
A way to write the building of a list, dict, or set in one line. Chapter 9.
computer vision
The field of making computers understand image content. Chapter 56.
concurrency
Handling several tasks in turn or at once so waiting time is not wasted. Chapter 35.
conda
A package and environment manager that also installs non-Python libraries. Chapter 17.
conditional (if/elif/else)
A statement that chooses a block of code based on whether a condition is true or false. Chapter 7.
confidence interval
A range of plausible values for a parameter, computed from a sample. Chapter 26.
confounder
A third variable that affects both variables and makes their relationship look misleading. Chapter 26.
confusion matrix
A table counting right and wrong guesses per class. Chapter 54.
context manager
An object for a with block that sets up and cleans up a resource automatically. Chapter 32.
copyleft and permissive
Copyleft requires derived works to stay open; permissive licences give more freedom. Chapter 63.
coroutine
An async def function that can pause while waiting and resume later. Chapter 35.
correlation
A measure of how much two variables move together; not proof of cause and effect. Chapter 26.
cosine similarity
A measure of how similar two vectors are through the angle between them. Chapters 28, 60.
CPython
The official Python implementation, written in C. Chapter 65.
cron
The periodic task scheduler on Linux and macOS. Chapter 41.
cross-validation
Testing a model repeatedly on rotating folds of the data. Chapters 27, 54.
CRS (coordinate reference system)
The rules that explain what coordinate numbers mean on the Earth’s surface. Chapter 38.
CSV
A text file format holding a table, with columns separated by commas. Chapter 14.
CUDA
NVIDIA’s platform that lets programs use its GPUs. Chapter 55.
Cython
A tool for compiling Python-like code into fast C extensions. Chapter 66.

D

dashboard
A page that sums up important numbers and charts at a glance. Chapter 42.
Dask
A parallel computing library for data that does not fit in memory. Chapter 53.
data drift
The spread of field data shifts away from the data used to train the model. Chapter 58.
data orchestration
Arranging, scheduling, and monitoring the steps of a data pipeline. Chapter 50.
data type
The kind of value, for example integer, decimal, text, or true/false. Chapter 5.
data validation
Checking incoming data against rules before it is used. Chapter 34.
database
Structured storage of data that can be queried. Chapter 24.
dataclass
A decorator that builds a data-holding class with minimal code. Chapter 16.
DataFrame
A pandas table of labelled rows and columns. Chapter 23.
debugging
The process of finding and fixing bugs. Chapter 12.
decision tree
A model made of a branching series of yes/no questions. Chapter 54.
decorator
A function that wraps another function to add behaviour, written with @. Chapter 31.
dependency
Another library a project needs in order to run. Chapter 17.
descriptor
An object that controls how an attribute is read and written. Chapter 64.
design pattern
A tested recipe for design problems that come up often. Chapter 46.
dict (dictionary)
A collection of key and value pairs, accessed by key. Chapter 9.
distributed system
A system whose work is divided among many processes or machines that communicate. Chapter 67.
Django
A full-featured Python web framework with many built-in parts. Chapter 36.
Docker
A tool for wrapping an application and its environment into containers. Chapter 48.
docstring
The sentence in triple quotes at the start of a function that explains its job. Chapter 10.
DOI
A permanent identifier that makes a dataset or paper easy to cite. Chapter 29.
DSM, DTM, and CHM
Surface model, ground model, and canopy height (CHM = DSM - DTM). Chapter 51.
dunder method
A method with double underscores at both ends, like __init__, called by Python automatically. Chapter 16.

E

effect size
The size of a difference or relationship, regardless of sample size. Chapter 26.
embedding
A vector of numbers that stands for the meaning of a text or object. Chapters 28, 60.
encapsulation
Hiding the inside of an object and opening it through controlled doors. Chapter 16.
encoding
The rule that turns letters into bytes in a file, for example UTF-8. Chapter 13.
.env file
A local settings file of environment variables; it must not enter Git. Chapter 34.
environment variable
A setting value outside the code, a safe place to keep secret keys. Chapter 21.
epoch and batch
An epoch is one pass over all training data; a batch is a group of examples per step. Chapter 55.
EPSG
The standard code number for a CRS, for example 4326. Chapter 38.
error
Python’s message that something needs fixing; read it from the last line. Chapter 4.
ETL and ELT
Data flows: extract, transform, load (ETL), or extract, load, then transform (ELT). Chapter 50.
event loop
The coordinator that takes turns running the coroutines that are ready. Chapter 35.
exception
The object Python raises when an error occurs, like ValueError. Chapter 12.
explainable AI (XAI)
Techniques that make a model’s decisions understandable to people. Chapter 58.

F

f-string
Text prefixed with f that inserts variable values inside curly braces. Chapter 5.
FastAPI
A modern web framework for APIs with Pydantic validation and automatic documentation. Chapter 36.
feature importance
A measure of how much each feature influences the model. Chapter 54.
features and label
Features are the model’s inputs; the label is the answer to be predicted. Chapter 54.
file
A named container of data in the computer’s storage, for example halo.py. Chapter 1.
file extension
The ending of a file name after the dot that marks its kind, such as .py or .csv. Chapter 1.
fine-tuning
Continuing to train an existing model on your own specific data. Chapter 61.
Flask
A lightweight Python web framework. Chapter 36.
float
A number with a decimal fraction, the float type; written with a point. Chapter 5.
flowchart
A drawing of steps with symbols and arrows that maps the flow of an algorithm. Chapter 2.
folder (directory)
A place that groups files; it can contain other files and other folders. Chapter 1.
for
A loop that takes the items of a collection one at a time. Chapter 8.
formatter
A tool that tidies the appearance of code automatically. Chapter 19.
free-threaded (no GIL)
A Python build without the GIL that lets threads truly run Python code at the same time. Chapter 47.
function
A named block of code written once and callable many times. Chapter 10.

G

garbage collector
A memory cleaner that discards unused objects that point at each other. Chapter 65.
generator
A function with yield that produces values one at a time and saves memory. Chapter 30.
generics
Type hints that name the type of a container’s contents, like list[int]. Chapter 33.
genetic algorithm
A solution search inspired by evolution: selection, crossover, and mutation. Chapter 57.
GeoPackage
A one-file format that stores many layers of spatial data. Chapter 38.
GeoPandas
The library that adds a geometry column to pandas tables. Chapter 38.
GeoTIFF
A raster file format that carries location information. Chapter 39.
GIL
A lock in CPython that lets only one thread run Python code at a time. Chapter 35.
GIS
Geographic information system: a way of storing and analysing data with locations. Chapter 38.
Git
A version-tracking system that records the history of file changes. Chapter 20.
GitHub
A site where Git repos are stored and worked on together. Chapter 20.
GitHub Actions
A GitHub service for running automatic workflows (such as tests) when changes arrive. Chapter 20.
.gitignore
A list file that tells Git which files must not be tracked. Chapter 20.
GLM (generalized linear model)
A family of linear models extended to count or category outcomes, for example Poisson. Chapter 27.
GPU
A graphics processor with thousands of small cores, good at simultaneous matrix multiplication. Chapter 55.
gradient boosting
Building trees one after another, each fixing the errors of the previous one. Chapter 54.
gradient descent
A way of training a model by shifting weights in the direction that lowers the error. Chapter 55.
group
The part of a pattern in parentheses that captures the matching piece of text. Chapter 22.
GUI
A graphical interface of windows, buttons, and input fields. Chapter 42.

H

hallucination
A model makes up an answer that sounds confident but is not true. Chapter 60.
hash and salt
A hash is a one-way fingerprint; a salt is a random addition that makes hashes hard to guess. Chapter 49.
health check
A light check that confirms a service is still healthy. Chapter 48.
HTTP
The communication rules between a browser or client program and a web server. Chapter 21.
HTTP status code
The server’s answer as a number, for example 200 for success and 404 for not found. Chapter 21.
hyperparameter
A model setting chosen before training, not learned from data. Chapter 57.
hypothesis test
A procedure for judging whether a difference in a sample is strong enough to count as real. Chapter 26.

I

I/O (input/output)
The process of reading data in and writing data out, for example to files. Chapter 13.
I/O-bound and CPU-bound
Tasks that mostly wait (I/O) differ from tasks that mostly compute (CPU). Chapter 35.
idempotent
A task that is safe to run again because the result is the same. Chapter 67.
image and container
An image is the packaging mold; a container is an image that is running. Chapter 48.
import
The statement that loads a module so its contents can be used. Chapter 11.
indentation
The spaces at the start of a line that mark a block of code in Python. Chapter 7.
index
The position number of an item in a list or text; it starts at 0. Chapter 9.
inheritance
A child class takes over data and methods from a parent class. Chapter 16.
integer
A number without a fraction, the int type. Chapter 5.
interactive mode (REPL)
A chat room with Python: you type one line of code and get the answer immediately. Chapter 4.
introspection
Inspecting objects, classes, and functions while the program runs. Chapter 64.
IoT (Internet of Things)
Physical sensor devices that send data over a network. Chapter 52.
IoU (intersection over union)
A measure of overlap between two boxes or areas: intersection divided by union. Chapter 56.
iterable
An object a for loop can go through, such as a list. Chapter 30.
iterator
An object that hands out its items one at a time when asked. Chapter 30.

J

JIT (just-in-time)
Translating code into machine code while the program runs; experimental in CPython 3.14. Chapter 47.
JSON
A standard text format for nested data of keys, values, and lists. Chapter 14.
Jupyter Notebook
An interactive workbook mixing code, results, and text. Chapter 25.

L

lambda
A tiny one-line function without a name. Chapter 10.
lazy evaluation
Delaying work until the result is truly needed so it can be optimised. Chapter 50.
LiDAR and point cloud
LiDAR measures distance with lasers; the result is a three-dimensional cloud of points. Chapter 51.
linter
A tool that inspects code and warns without running it. Chapter 19.
list
An ordered collection that can be changed, written with square brackets. Chapter 9.
LLM (large language model)
An AI model trained on huge amounts of text so it can write and answer. Chapter 60.
lock file
A file that pins the exact version of every dependency, for example uv.lock. Chapter 17.
logging
The built-in module for recording program events with levels of importance. Chapter 19.
logical operator
and, or, and not: they combine or flip true/false values. Chapter 6.
logistic regression
Regression for yes/no outcomes that yields a probability. Chapter 27.
loop
A block of code run again and again, with for or while. Chapter 8.
LoRA and QLoRA
Memory-saving fine-tuning that trains only a few added parameters. Chapter 61.

M

machine learning
A way of making a computer learn patterns from example data instead of hand-written rules. Chapters 25, 54.
mAP (mean average precision)
The average score of object detection quality across all classes. Chapter 56.
match/case
A statement that matches a value against several patterns. Chapter 7.
matplotlib
The basic library for drawing charts in Python. Chapter 25.
MCP (Model Context Protocol)
A standard protocol for connecting models to tools and data sources. Chapter 62.
median
The middle value after sorting; resistant to extreme values. Chapter 26.
merge
Combining the histories of two branches into one. Chapter 20.
merge conflict
The situation where two branches changed the same line so a person must choose. Chapter 20.
metaclass
A class that creates classes. Chapter 64.
metaprogramming
Code that reads or writes other code, or shapes classes while the program runs. Chapter 64.
method
A function that belongs to a class. Chapter 16.
MicroPython
A slimmed-down Python for microcontrollers. Chapter 52.
MLflow
A tool for tracking experiments and models. Chapter 58.
MLOps
The practice of managing a model’s life cycle: experiments, versions, release, monitoring. Chapter 58.
model card
A fact sheet for a model: purpose, data, limits, and risks. Chapter 58.
module
A single .py file of code that can be used through import. Chapter 11.
Monte Carlo
A technique of drawing random numbers many times to estimate a result. Chapter 27.
MQTT
A light publish-subscribe protocol for sending sensor data through a broker. Chapter 52.
multicollinearity
When several explanatory variables are tightly related so the coefficients become unstable. Chapter 27.

N

NaN
The “not a number” value that marks empty or missing data. Chapter 23.
NDVI
A vegetation greenness index from the red and near-infrared bands. Chapter 39.
neural network
A model of layered artificial neurons that learns by adjusting weights. Chapter 55.
None
Python’s special value meaning “no value”. Chapter 10.
normal distribution
A bell-shaped spread symmetric around the average. Chapter 26.
NSGA-II
A multi-objective evolutionary algorithm that ranks solutions by domination and crowding distance. Chapter 57.
numpy
The library for fast arrays of numbers. Chapter 23.

O

object
A product of a class; it carries data and methods. Chapter 16.
object detection
Finding the location and kind of objects in an image with bounding boxes. Chapter 56.
observability
The ability to understand a system’s state from its logs, metrics, and traces. Chapter 48.
OLS (ordinary least squares)
The most common way to fit a regression line by minimising the sum of squared differences. Chapter 27.
OOP (object-oriented programming)
A way of building programs as objects that carry both data and the functions that go with it. Chapter 16.
open source
Software whose code is open and may be studied and contributed to. Chapter 68.
OpenCV
A popular library for processing images and video. Chapter 56.
operator
A symbol that does work on values, like +, ==, or and. Chapter 6.
ORM
The technique of treating table rows as Python objects. Chapter 24.
overfitting
A model memorises the training data and does badly on new data. Chapter 54.
overlay
Stacking two map layers to get their intersection or union. Chapter 38.

P

p-value
The chance of seeing a result that strong if the null hypothesis is true; not the chance your hypothesis is true. Chapter 26.
package
A folder of several modules that can be imported together. Chapter 11.
pandas
The library for labelled tables of data. Chapter 23.
parameter
A name in the parentheses of a function definition that receives input. Chapter 10.
Pareto front
The set of solutions that cannot improve on one goal without hurting another. Chapter 57.
Parquet
A columnar file format that is small and fast to read in part. Chapter 50.
partial dependence
A plot of the average effect of one feature on the prediction. Chapter 58.
path traversal
An attack through input, for example ../ to escape the allowed folder. Chapter 49.
pathlib
The built-in module for handling file and folder addresses as Path objects. Chapters 11, 13.
PEP 8
The commonly accepted style guide for writing Python code. Chapter 19.
Personal Data Protection Law (UU PDP)
Indonesia’s Personal Data Protection Law. Chapter 63.
pickle
A Python module that saves any object; unsafe for files from untrusted sources. Chapter 14.
pip
Python’s built-in program for installing libraries from PyPI. Chapter 11.
pixel
A single small cell in a raster that holds one value. Chapter 39.
polymorphism
Different objects respond to a method call of the same name in their own ways. Chapter 16.
portfolio
A selected collection of projects that shows what you can do. Chapter 68.
ports and adapters
A design that separates the application core from outside tools through swappable interfaces. Chapter 46.
PostGIS
The PostgreSQL extension for storing and querying spatial data. Chapter 40.
PostgreSQL
A powerful, open server database system. Chapter 24.
precision and recall
Precision: how many positive guesses were right. Recall: how many true positives were found. Chapter 56.
process
A running program with its own memory; many processes can use many CPU cores. Chapter 35.
profiling
Measuring which parts of a program use time or memory. Chapter 44.
program
A list of instructions the computer carries out step by step. Chapter 1.
programming language
A written language people use to give instructions to a computer. Chapter 1.
projection
A way of mapping the curved Earth onto a flat plane. Chapter 38.
prompt
The mark in the terminal that shows the computer is ready for a command. Chapter 3.
prompt
The instruction and context text you send to a language model. Chapter 60.
property
A decorator that makes a method look like an attribute, usually for checking. Chapter 16.
pseudocode
Algorithm steps written in tidy everyday language, not yet real code. Chapter 2.
pull request
A request to merge a branch into the main line, with review. Chapter 20.
Pydantic
A library that validates and converts incoming data based on typed models. Chapter 34.
PyPI
The online warehouse where Python libraries are shared and downloaded. Chapter 11.
pyproject.toml
The standard configuration file of a Python project (name, version, dependencies, tools). Chapter 17.
PyQGIS
The Python interface for automating QGIS. Chapter 40.
pytest
The most popular testing framework in Python. Chapter 18.
Python
An easy-to-read, versatile, free, and open programming language; the main language of this book. Chapter 4.
Python install manager
The official manager for installing and choosing Python versions on Windows; its command is py. Chapter 4.
PyTorch
A deep learning library built on tensors and autograd. Chapter 55.

Q

QGIS
A free, open desktop GIS application. Chapter 40.
qualitative methods
Ways of researching that answer “why” and “how” through stories and words. Chapter 28.
quantization
Reducing the precision of a model’s weights so it fits in less memory. Chapter 61.

R

R-CNN
A family of two-stage detectors: propose regions, then classify. Chapter 56.
R-squared
The share of variation in the outcome that the model explains, from 0 to 1. Chapter 27.
RAG
Answering by first searching for related documents, then giving them to the model. Chapter 60.
RAM (memory)
The computer’s temporary workspace; its contents vanish when the computer is switched off. Chapter 1.
random forest
A collection of many decision trees that pick the answer by voting. Chapter 54.
range
A function that produces a sequence of numbers for a loop. Chapter 8.
raster data
Data in a grid of pixels, such as satellite images. Chapter 39.
README
A project’s introductory file: what it is, how to install, how to use. Chapter 68.
Redis
A fast in-memory data store, often used as a broker. Chapter 67.
reference counting
How Python counts the names pointing at an object and frees it at zero. Chapter 65.
regex (regular expression)
A text pattern for searching or matching pieces of text. Chapter 22.
regression
Building an equation that links one outcome to one or more causes. Chapter 27.
regularization (Ridge and Lasso)
Penalising large coefficients so the model does not memorise the data too much. Chapter 27.
remote
A copy of the repo on another server, for example on GitHub. Chapter 20.
repository
A project folder whose history Git tracks. Chapter 20.
reproducible research
Research whose data and code others can run to get the same numbers. Chapter 29.
residual
The difference between the actual value and the model’s predicted value. Chapter 27.
REST API
A web API that uses addresses and HTTP methods to handle data resources. Chapter 36.
return
The statement that sends a function’s result back to its caller. Chapter 10.

S

schema
The agreed design of columns and types for a dataset. Chapter 14.
scikit-learn
The classic machine learning library in Python. Chapter 25.
scope
The region where a variable can be seen, for example local inside a function. Chapter 10.
segmentation
Labelling every pixel of an image by its class. Chapter 56.
self
The name for the object itself inside a method. Chapter 16.
serialization
Turning data in memory into text or bytes that can be stored or sent. Chapter 14.
set
A collection of unique values with no order. Chapter 9.
SHAP
A method that splits each feature’s contribution to one prediction. Chapter 58.
Shapefile
An older vector data format made of five or more files. Chapter 38.
shell
The program inside the terminal that reads and runs the commands you type, for example PowerShell or bash. Chapter 3.
slicing
Taking part of a list or text with [start:end]. Chapter 9.
software license
The official terms that decide how code may be used, changed, and shared. Chapter 63.
SOLID
Five object-oriented design principles that make code easier to change. Chapter 46.
spatial index (GiST)
A helper structure that speeds up searches by location. Chapter 40.
spatial join
Joining tables by location relationship rather than by column value. Chapter 38.
SQL
The language for asking for and changing data in a database. Chapter 24.
SQL injection
An attack that slips SQL commands in through input; prevented with parameters. Chapter 24.
SQLAlchemy
A Python library for working with databases, with Core and ORM. Chapter 24.
SQLite
A small database that lives entirely in one file. Chapter 24.
STAC
A catalogue standard for finding satellite images and geospatial data. Chapter 51.
staging area
The area where changes are gathered with git add before committing. Chapter 20.
standard deviation
A measure of how far data spreads from its average. Chapter 26.
standard library
The collection of modules that come installed with Python. Chapter 11.
Streamlit
A library for making interactive data apps with just a Python script. Chapter 43.
string
A sequence of characters wrapped in quotation marks, like "Teak". Chapter 5.
subinterpreter
Several separate Python interpreters inside one process. Chapter 47.
subprocess
The module for running other programs from Python and reading their results. Chapter 41.
SyntaxError
An error because Python’s writing rules were broken, for example an unpaired quotation mark. Chapter 4.

T

task graph
The map of computation steps and their dependencies that Dask builds before running. Chapter 53.
task queue
A waiting list of jobs that workers pick up and carry out. Chapter 67.
template string
A new string kind (t"...") that keeps text parts and values separate. Chapter 47.
tensor
A multidimensional array of numbers that can live on a CPU or GPU. Chapter 55.
terminal
A window where you type text commands to the computer. Chapter 3.
TF-IDF
A word weight that is high when a word is frequent in one document but rare in others. Chapter 28.
thematic coding
Labelling pieces of interview text with codes to find themes. Chapter 28.
thread
A line of work inside one process that shares memory. Chapter 35.
timeout
The time limit for waiting; past it the request is cancelled. Chapter 21.
token and tokenization
A token is a piece of text a model reads; tokenization splits text into tokens. Chapter 61.
tool
A function an agent may call: the model asks your program to run it (tool use or function calling), then uses the result. Chapter 62.
topic modeling
An automatic technique for finding groups of words that form topics across many documents. Chapter 28.
traceback
The step-by-step report Python shows on an error, pointing to where it happened. Chapter 12.
training and test data
Two parts of the data: one to train the model, one to test it on data it has never seen. Chapter 25.
transaction
A group of data changes that either all succeed or are all cancelled. Chapter 24.
transformer
A neural network architecture based on attention that underlies modern language models. Chapter 61.
Trusted Publishing
A way of publishing packages to PyPI from an automated workflow without storing a password. Chapter 45.
try/except
A block that tries some code and catches the error if one occurs. Chapter 12.
tuple
An ordered collection that cannot be changed, written with round brackets. Chapter 9.
type hint
A note of types in code that Python does not enforce but checking tools read. Chapter 33.

U

unpacking
Splitting the contents of a tuple or list into several variables at once. Chapter 9.
UTF-8
A text encoding that can store nearly every letter and symbol in the world; the safe choice. Chapter 13.
uv
A fast new-generation Python package and project manager that can also install Python. Chapter 17.

V

variable
A name that points to a value the program keeps. Chapter 5.
vector data
Map data made of points, lines, and polygons. Chapter 38.
vector database
A store that finds data by the similarity of embedding vectors. Chapter 60.
vectorization
Applying an operation to a whole array at once, without a manual loop. Chapter 23.
venv
Python’s built-in module for creating virtual environments. Chapter 17.
virtual environment
A separate room holding the Python and libraries of one project. Chapter 17.
VRAM
The GPU’s own dedicated memory; the main limit on the size of model you can run. Chapter 55.

W

web scraping
Collecting data from web pages automatically, with ethics and site rules. Chapter 41.
WebAssembly
A code format that runs in the browser at near-native speed. Chapter 37.
wheel and sdist
Two forms of a Python package file: wheel is ready to install, sdist holds source code. Chapter 45.
while
A loop that keeps running while its condition is true. Chapter 8.
worker
A process that takes tasks from the queue and does them. Chapter 67.

X

xarray
A library for labelled multidimensional arrays such as imagery over time. Chapter 51.

Y

yield
The statement that hands over one value and pauses the generator function. Chapter 30.
YOLO
A family of fast one-stage object detectors. Chapter 56.

Z

zonal statistics
Summarising raster values inside a zone, for example per plot. Chapter 39.

Where to go next

For reading paths by goal (for example GIS or data/AI), see Appendix E. For further reading on each topic, see Appendix F.