Pandas Series
Cheatsheet

import pandas as pd

s = pd.Series([10, 20, 30],
             index=['a','b','c'],
             name='scores')

s = pd.Series({'a': 10, 'b': 20, 'c': 30})

s = pd.Series(5, index=['a','b','c'])

import numpy as np
s = pd.Series(np.array([1.1, 2.2, 3.3]))

s.values   # NumPy array underneath
s.index    # Index(['a','b','c'])
s.dtype    # dtype('int64')
s.name     # 'scores'
s.shape    # (3,) — always 1D
s.size     # 3 — same as len(s)
s.ndim     # 1 — always for Series
s.empty    # True if len == 0
s.hasnans  # True if any NaN

s['a']              # → 10
s[['a','c']]         # → Series
s.loc['a']          # explicit label
s.loc['a':'c']      # label slice (inclusive)
s.at['a']           # fastest single

s.iloc[0]           # first → 10
s.iloc[-1]          # last element
s.iloc[0:2]         # first 2 (exclusive end)
s.iloc[[0,2]]       # 1st and 3rd
s.iat[0]            # fastest single

s[s > 20]
s[(s>10) & (s<40)]
s[s.isin([10,30])]

s + 10          # add to every element
s * 2           # multiply each
s ** 2          # square each
np.sqrt(s)     # numpy ufunc works!
np.log(s)      # natural log
s1 + s2        # aligns by index

s.sum()   s.mean()  s.median()
s.min()   s.max()   s.std()
s.var()   s.prod()  s.sem()

s.isna()         # True where NaN
s.notna()        # True where not NaN
s.isna().sum()  # count of NaNs
s.isna().any()  # any missing?
s.isna().all()  # all missing?

s.dropna()           # remove NaN rows
s.fillna(0)           # replace with 0
s.fillna(s.mean())   # impute with mean
s.ffill()             # forward fill
s.bfill()             # backward fill
s.interpolate()       # linear fill
s.interpolate(method='cubic')

s.str.upper()      # "hello" → "HELLO"
s.str.lower()      # "HELLO" → "hello"
s.str.title()      # "hello world" → "Hello World"
s.str.strip()      # remove surrounding spaces
s.str.lstrip()     # left only
s.str.rstrip()     # right only
s.str.len()        # length of each string

s.str.contains('py')       # bool
s.str.startswith('A')     # bool
s.str.endswith('ing')    # bool
s.str.match(r'^\d+')     # regex at start
s.str.fullmatch(r'\d+') # full match
s.str.count('l')        # count occurrences

s.str.split(',')              # → list per row
s.str.split(',', expand=True) # → DataFrame
s.str.extract(r'(\d+)')       # capture group
s.str.extractall(r'(\d+)')   # all matches
s.str[0:3]                    # slice chars

s.str.replace('a','A')
s.str.replace(r'\d','X', regex=True)

s = pd.to_datetime([
  '2024-01-15', '2024-06-30'])

s.dt.year         # [2024, 2024]
s.dt.month        # [1, 6]
s.dt.day          # [15, 30]
s.dt.hour         # hour of day
s.dt.minute       # minute
s.dt.dayofweek   # 0=Mon … 6=Sun
s.dt.day_name()  # 'Monday'…
s.dt.quarter     # 1–4

s.dt.is_month_end    # bool
s.dt.is_month_start  # bool
s.dt.is_leap_year    # bool
s.dt.is_quarter_end  # bool

# Timezone ops
s.dt.tz_localize('UTC')
s.dt.tz_convert('Asia/Kolkata')

# Formatting
s.dt.strftime('%d-%b-%Y')

s.describe()        # full summary
s.mean()   s.median()
s.std()    s.var()
s.min()    s.max()
s.skew()             # distribution skew
s.kurt()             # kurtosis (peakedness)
s.sem()              # standard error of mean
s.mad()              # mean absolute deviation
s.quantile(0.25)    # 25th percentile
s.quantile([.25,.5,.75])

s.value_counts()             # freq table
s.value_counts(normalize=True) # proportions
s.value_counts(dropna=False)  # include NaN
s.value_counts(bins=4)        # histogram bins
s.unique()                    # array of uniques
s.nunique()                   # count of uniques
s.mode()                      # most common value(s)
s.mode()[0]                   # top mode

s.rank()                 # 1-based rank
s.rank(ascending=False)  # highest = 1
s.rank(method='dense')  # no gaps
s.rank(method='min')    # tie = min rank
s.rank(pct=True)        # 0–1 percentile

s.cumsum()   # running total
s.cumprod()  # running product
s.cummax()   # running maximum
s.cummin()   # running minimum

s.rolling(3).mean()   # 3-point moving avg
s.rolling(3).sum()
s.rolling(3).min()
s.rolling(3).std()
s.rolling(3).max()
s.rolling(3, min_periods=1).mean()

s.expanding().mean()  # growing average
s.expanding().sum()   # = cumsum()
s.expanding().max()
s.expanding(2).std()  # min 2 periods

s.between(10, 30)      # 10 ≤ x ≤ 30 → bool
s.isin([10, 30])       # exact match → bool
s.clip(lower=10, upper=30) # clamp range
s.where(s > 20)        # keep, else NaN
s.mask(s > 20)         # NaN where True
s.gt(20)  s.lt(20)      # .gt .lt .ge .le .eq

s.map({1:'low', 2:'mid', 3:'high'})
s.map(lambda x: x * 10)
s.map(str)  # convert each to string

s.replace(10, 99)            # exact swap
s.replace({10:99, 20:88})   # multi-swap
s.replace(r'\d+', 'N', regex=True)
s.apply(lambda x: x ** 2)    # custom fn
s.apply(np.log)               # numpy fn

s.duplicated()           # bool mask
s.duplicated(keep='last')
s.duplicated(keep=False)  # mark ALL dupes
s.drop_duplicates()        # remove dupes
s.drop_duplicates(keep='last')
s[~s.duplicated()]         # filter unique

s.shift(1)       # lag by 1 (prev row)
s.shift(-1)      # lead by 1 (next row)
s.diff()         # s − s.shift(1)
s.diff(2)        # 2-period difference
s.pct_change()   # (s−prev)/prev → %
s.pct_change(7) # 7-day % change

s.index = ['x','y','z']   # direct set
s.rename({'a':'A'})    # rename labels
s.rename(index=str.upper) # via function
s.reset_index()           # → DataFrame
s.reset_index(drop=True)  # keep as Series
s.set_axis(['p','q','r'])

s.reindex(['a','b','d'])    # new index
s.reindex(fill_value=0)      # fill missing
s.reindex(method='ffill')   # forward fill
s1.align(s2)                  # align two Series

s.sort_values()              # asc by default
s.sort_values(ascending=False)
s.sort_values(na_position='first')
s.nlargest(3)    # top 3 (fast!)
s.nsmallest(3)  # bottom 3 (fast!)

s.sort_index()             # A→Z or 0→n
s.sort_index(ascending=False)
s.sort_index(na_position='last')

pd.concat([s1, s2])          # stack rows
pd.concat([s1, s2], axis=1)  # side by side
pd.concat([s1, s2],
          ignore_index=True)   # reindex 0,1,2…
pd.concat([s1, s2], keys=['a','b']) # multi-index

s1.combine(s2, max)          # element-wise max
s1.combine_first(s2)          # fill s1's NaN
                              # with values from s2

s.astype(float)           # to float64
s.astype('Int64')         # nullable int
s.astype(str)             # to string (object)
s.astype('category')      # memory efficient
s.astype('datetime64[ns]')
pd.to_numeric(s)          # safe conversion
pd.to_numeric(s, errors='coerce') # NaN on fail

s.to_list()      # Python list
s.to_dict()      # {index: value}
s.to_frame()     # 1-column DataFrame
s.to_numpy()     # NumPy array
s.to_csv('f.csv')
s.to_json()      # JSON string
s.to_clipboard() # copy to clipboard