Using uv? The #Python version is stored in .python-version. > cat .python-version 3.14 Change it with > uv python pin 3.15 # not out yet! But: > uv python pin 3.13 uv gives an error: 3.13 is incompatible with requires-python in pyproject.toml of '>=3.14.'

Is the US falling behind China in research? The Leiden Institute recently published its research rankings — and in the latest Bamboo Weekly, we used Claude Code to write #Python #Pandas to analyze this data set. Learn more about AI, data analysis, and the future of research: bambooweekly.com

Want to get the unique values from a #Python list? Use a set! import random numbers = [random.randint(0, 10) for i in range(20)] set(numbers) # returns a set of unique numbers {*numbers} # same result, but uses unpacking syntax Of course, list elements must be hashable!

Software engineers at major companies are using AI to write production code. Right now. Want to join them? I'm running two AI-powered workshops (#Python and #Pandas) where you'll learn about prompts, output, and evaluating AI output. Details: buff.ly/aGvI1bd

For over a year, my answer to "How is AI affecting your training business?" was glib: "The biggest effect? Everyone asks me that question." I'm not saying that anymore. AI is now part of coding. This means software engineers need more skills to thrive: lerner.co.il/2026/01/21/w...

🎉 Pandas 3 is out! As of a short while ago, installing #Python #Pandas from PyPI, with either pip or uv, will give you the latest version. What's new? What has changed? 🤔 I've got a whole YouTube playlist to tell you: www.youtube.com/playlist?lis...

Are US universities falling behind? In the latest Bamboo Weekly, I ask 6 questions from the latest data. But don't write the #Python #Pandas queries yourself — rather, ask Claude Code to do it for you! Level up your data skills every Wednesday: bambooweekly.com

How much memory does a text column in your #Python #Pandas data frame use? Check: df['x'].memory_usage() But this is likely a huge underestimate! It sums the pointer sizes, not the string sizes. Instead, say: df['x'].memory_usage(deep=True) Or: df.info(memory_usage='deep')

Thanks to the people who have not only bought the second edition of Python Workout (my book with 200 exercises), but who have taken the time to write reviews on Amazon. And wow, they are saying some really, *really* nice things! Check out what they've said: www.amazon.com/Python-Worko...?

Want to turn an online CSV file into a #Python #Pandas data frame? Just pass a URL to read_csv: df = pd.read_csv('https://mydata.com/data.csv') options such as usecols and index_col work! This works on all read_* methods, including read_excel, read_json, and read_parquet.

Had enough #Python classes and exercises? It's time to dirty your hands, creating a personal project! Join HOPPy (Hands-On Projects in Python), where you create a data dashboard using Marimo under my mentorship. Questions? Join my Jan 26 info session: us02web.zoom.us/webinar/regi...

After plotting a #Python #Pandas data frame in Jupyter, how do you save it to a file? # Grab the axes ax = df.plot.scatter(x='trip_distance', y='total_amount') # Get the figure fig = ax.get_figure() # Filename extension determines format fig.savefig('/tmp/myplot.png')

If your #Python #Pandas dataframe has a datetime index, use a slice to retrieve rows in a certain time period: taxi_df.loc['2025-11-01':'2025-11-03'] But if the rows are unsorted? You get a KeyError. Solution: taxi_df.sort_index().loc['2025-11-01':'2025-11-03']

If a text column #Python #Pandas data frame uses lots of memory and repeats, make it a category: df['x'].memory_usage(deep=True) # result: 8,093,417 df['x'] = df['x'].astype('category') df['x'].memory_usage(deep=True) # result: 154,853 A 99% savings!

Sorting a #Python dict by value? Instead of: sorted(d.items(), key=lambda t: t[1]) Use itemgetter. It's cleaner AND ~25% faster: from operator import itemgetter sorted(d.items(), key=itemgetter(1)) Sort by value, then by key: sorted(d.items(), key=itemgetter(1, 0)) # value,key

Write to a file in #Python with open(filename, 'w'). What if filename already exists? Its contents are gone. (I hope you have backups!) Instead, try: open(filename, 'x') If filename already exists, 'x' raises an exception. For new files, 'x' and 'w' are the same.

How much oil does Venezuela export vs. Saudi Arabia? Which country's exports are more highly correlated with oil prices? And how much has oil consumption changed in Venezuela in the last 15 years? We calculate it all, using #Python #Pandas, in the latest Bamboo Weekly!

To sort a #Python #Pandas data frame by column x: df.sort_values('x') By x and y (ascending): df.sort_values(['x', 'y']) By x and y, both descending: df.sort_values(['x', 'y'], ascending=False) Ascending x, descending y: df.sort_values(['x', 'y'], ascending=[True, False])

In the latest Bamboo Weekly, we use #Python #Pandas to analyze Venezuelan oil production and consumption — including comparing their output with Saudi Arabia. Improve your data skills with real data and current events! New questions every Wednesday. Solutions every Thursday.

A #Python #Pandas rule of thumb: Wherever you can use a column name (string), you can use multiple column names (a list of strings): df.set_index('x') df.set_index(['x', 'y']) df.groupby('last') df.groupby(['last', 'first']) df['x'] df[['x', 'y']] # Notice nested []

Retrieve rows of a #Python #Pandas data frame with the 6 largest values of column x: df.sort_values('x').tail(6) Even better: df.nlargest(n=6, columns='x') Why is it better? On a 3m row, 680 MB data frame, sort/tail took 622 ms. nlargest? Only 84 ms — 7x faster!

Why is #Python's join() a STRING method, not a list method? x = 'this is a test'.split() '*'.join(x) ':::'.join(x) Not: x.join('*') Why not? The argument can be any string-returning iterable: '*'.join('abcde') '*'.join(str(x) for x in range(5)) '*'.join(open('filename.txt'))

Need specific column types in #Python #Pandas? Use select_dtypes: df.select_dtypes(include='int64') df.select_dtypes(include=['int64', 'float64']) df.select_dtypes(exclude='int64') df.select_dtypes(exclude='number') # all numbers df.select_dtypes(include='str') # Pandas 3 str

Use parentheses to split long #Python code across lines: if x == 10 and y == 20: print('Yes!') # ☹️ But with parentheses, Python sees it as one line: if (x == 10 and y == 20): # 🙂 print('Yes!') Or in comprehensions… [x*5 for x in range(10) if x % 2]

#Python Jupyter memory quirk: Deleting variables doesn't always free memory! # cell 1 df = [YOUR BIG DATAFRAME] df # View it + stores in Out[1] # cell 2 del(df) # df deleted, but Out[1] still has it! Use df.head() instead of df to keep big values out of Jupyter's Out dict.

New York began congestion pricing in 2025. Also in 2025? Taxis and buses moved faster, and more people took public transportation. We can prove it with the data — and using #Python #Pandas. Challenge yourself with new, weekly real-world data problems: bambooweekly.com

Playing NYT Spelling Bee? Use #Python to see if your word only uses allowed letters (ignoring length + center): letters = set('ygoanik') # From Mon, Jan 5 With sets, <= checks "subset or equal": set('again') <= letters # True set('garage') <= letters # False (r + e)

Congestion pricing went into effect in New York one year ago. How did it affect traffic? Are buses faster than before? The latest Bamboo Weekly asks you to answer these questions (and more) with Pandas. Level up your data-analysis skills at bambooweekly.com

Create a #Python dict with keys a/b/c and values 0: d = dict.fromkeys('abc', 0) But don't do this, since the value will be shared and mutable: d = dict.fromkeys('abc', []) d['a'].append('x') d['b'].append('y') print(d) {'a': ['x', 'y'], 'b': ['x', 'y'], 'c': ['x', 'y']} 🤯

"round" rounds #Python numbers: round(1.3) # 1 round(1.6) # 2 round(1.61, 1) # 1.6 round(1.66, 1) # 1.7 Remember, .5 goes to the nearest *even* value: round(1.5) # 2 round(2.5) # 2, yes 2! 🤯 round(3.5) # 4 round(1.65, 1) # 1.6 round(2.65, 1) # 2.6 🤯 round(3.65, 1) # 3.6

Want to reverse a #Python sequence (string/list/tuple)? Use a 3-argument slice: s[::-1] This works because: - Empty start means "from the start" - Empty end means "through the end" - Step size of -1 means "go back 1 each time" This returns a new value, from s's end to its start.

#Python is all about consistency. For example, sequences (string/list/tuple) all support: s[i] # get 1 item s[start:finish] # get slice 'a' in s # search s.index('a') # where is a? s.count('a') # how many as? for item in s: # iterable print(item)

How can you tell if a #Python datetime is a weekend (i.e., Saturday or Sunday)? Invoke weekday(). It returns an int; Mon is 0, Sat is 5 and Sun is 6. So: today = datetime.datetime.now() if today.weekday() > 4: print('Weekend!') else: print('Not yet the weekend. :-(')"

Consider this #Python: >>> x = 100000 >>> y = 100000 >>> x is y False Surprised? That's because "is" is the wrong comparison: • == asks: are these the same value? • is asks: are these the same object? You almost never care about question #2. And so, should rarely use "is".

Want to make #Python numbers more readable? Use _ between digits (not at the start or end): x = 1_000_000 y = 2_000 z = 1_000.234_567

The new year is off to a great start: My printed copies of Python Workout just arrived!

Wondering how Pandas 2 looks vs. Pandas 3? In my latest video, I compare the spread of the flu in the US using both versions, comparing dtypes, memory usage, and query styles on real-life CDC data: youtu.be/1OqOM0FmFF0?...

In the final Bamboo Weekly for 2025, we ask: • Does PyCon US affect PyPI downloads? • Is uv gaining traction? • How fast are people adopting Python 3.14? Check it out: bambooweekly.com

Time zones are a pain to deal with! #Pandas 3 makes things a bit better by using the now-standard #Python zoneinfo module. In this video, I show you what has changed — but also where/why time zones can cause so many headaches, and what Pandas 3 changed. Check it out: youtu.be/Zekos7T2KI4?...

Nervous about upgrading #Pandas, because you don't know what'll break? My latest video explains how Pandas is standardizing warnings, to make it clearer, now and in the future: youtu.be/lwBCX2FuB-4?...

Are you ready for #Python #Pandas 3.0? Find out with my Pandas 3 playlist, youtu.be/TCSjgvtO714?.... New additions: • Changes to sorted pd.concat and time series: youtu.be/2DVov1OfCFI?... • Changed resample offsets: youtu.be/rZutMLQmWcY?...

What proportion of Kalshi's contracts are about sports? In the latest Bamboo Weekly, we use #Pandas to look into this popular prediction market, calculating the number of contracts in each category and then plotting them. Get real-world Pandas practice every Wednesday: bambooweekly.com

Do you use categories in Pandas? And do you use groupby? #Pandas 3 will change the default behavior, from observed=False to observed=True. In my latest video, I explain what this means, and what you'll need to do when upgrading: youtu.be/bmf7crcJm9k

How much of the Kalshi prediction market's contracts are about sports? And did court rulings last October help their business? In the latest Bamboo Weekly, we use Pandas to analyze data about Kalshi's markets. Level up your data-analysis skills every Wednesday: bambooweekly.com

Does your data include dates and times? In #Python #Pandas 3, datetimes are stored in microseconds (us) by default, not ns. My latest video explains what's going on: www.youtube.com/watch?v=poQo... Catch this, and my entire "Prep for Pandas 3" playlist: youtu.be/poQo-V-3Gp4?...

What did I do in 2025? In a nutshell: Lots of teaching #Python and #Pandas, writing books and newsletters, recording videos, and traveling -- with lots more planned for 2026! You can see my writeup here: lerner.co.il/2025/12/23/r...

#Python #Pandas 3 is coming soon — and it's different from Pandas 2. I just released two new videos: • inplace=True now sometimes returns a value: youtu.be/VySIM5CW8Kw • SettingWithCopyWarning has gone away, thanks to copy on write: youtu.be/wITIN2DOXtU What questions do you have about Pandas 3?

#Python #Pandas 3 is coming out soon — and a lot of things are changing. To help you prepare, I've got new videos on my YouTube channel. Get them all at my new Pandas 3 playlist: www.youtube.com/playlist?lis...

How many people who feel sick actually have the flu? In the latest Bamboo Weekly, we compared numbers from last year and this year for each HHS region. Oh, and we compare memory, speed, and syntax of #Pandas 2 vs. 3. Level up your Pandas with real-world questions and data sets: bambooweekly.com

In the latest Bamboo Weekly, we examine the recent spread of flu in the US with data from the CDC. Better yet: We compare Pandas 2 and Pandas 3 for speed, memory, and style. You'll feel better after solving these real-world data problems! Check them out at bambooweekly.com