Creating, Reading, Writing
df = pd.DataFrame({"col0": [val0, val1], "col1": [val0, val1]}, index=[0, 1]) |
Create a dataframe |
series = pd.Series(["val0", "val1", "val2"], index=[0, 1, 2, 3], name="name") |
Create a series |
read = pd.read_csv("../folder/folder/file.csv", index_col=0) |
Read a csv |
save.to_csv("file.csv") |
Save an existing dataframe as a csv |
Indexing, Selecting, Assigning
table.head |
Show first 5 rows of a dataframe |
table["col"] |
Select the col from table |
table.col.iloc[0] |
Select 1st value of a col from table |
table.iloc[0] |
Select 1st row of data from table |
table.col.iloc[:10] |
Select 1st 10 values from col in table (index-based select) |
table.col.loc[:10] |
Select 1st 10 values from col in table (label-based select) |
table.loc[indices, cols] |
Select certain rows from certain cols |
table[table.col == 'val'] |
Select cols have a certain val (conditional select) |
table.col.isin(['val1,' 'val2']) |
Select cols have certain vals (conditional select) |
Summary Functions & Maps
table.col.describe() |
Get high-lvl summary of given col's attributes |
table.col.mean() |
Get mean of a col with numerical vals |
table.col.unique() |
Get each unique val of a col w/ no dupes |
table.col.value_counts() |
Get frequency of each val in col |
table.col.map(lambda p: p - s) |
Map function to remap a Series of point vals (p) by using a transformation (p-s) -> returns new Series |
table.apply(func, axis='columns') |
Apply function to transform entire df by calling custom method (func taking a row) on each row |
|
|
Grouping & Sorting
table.groupby('col').col.count() |
Group data w/ same vals in the given col -> count frequency of given col (same as value_counts()) |
table.groupby('col').size() |
Same as above |
table.groupby('col').apply(lambda df: df.title.iloc[0]) |
Select name (title) of the 1st thing in col |
table.col.idxmax() |
Get index of max val in col |
table.groupby(['col0']).col1.agg([f1, f2, f3]) |
agg() runs diff. funcs. simultaneously on a df |
table.groupby(['col0', 'col1']).col2.agg([len]) |
Multi-index output has tiered structure. Require 2 levels of labels to retrieve a val |
df.reset_index() |
Muti-index method used to converting back to regular index |
df.sort_values(by='col') |
Sort rows of data by vals in col (ascending) |
df.sort_values(by='col', ascending=False) |
Sort rows of data by vals in col (descending) |
df.sort_values(by=['col0', 'col1']) |
Sort rows by more than 1 col at a time |
df.sort_index() |
Sort rows by index (default order; ascending) |
Data Types & Missing Values
table.col.dtype |
Get data type of a col |
table.dtypes |
Get data types of each col in table |
table.col.astype('datatype') |
Convert col to datatype if allowed (e,g, int64 -> float64) |
table.index.dtype |
Number indices are int64 |
table[pd.isnull(table.col)] |
Select NaN entries in a col |
table.col.fillna("filler") |
Replace all NaN vals in a col with a sentinel val ("Unknown", "Undisclosed", "Invalid") or non-null val |
table.col.replace("init_val", "new_val") |
Replace, in col, all existing vals with new_vals |
Renaming & Combining
table.rename(columns={'init': 'new'}) |
Rename col or index col names |
table.rename(index={0: 'firstEntry', 1: 'secondEntry}) |
Rename index or col vals by specifying an index or col param |
table.rename_axis("name", axis='rows').rename_axis("name1", axis='columns') |
Rename row index &/or col index |
pd.concat(list, of, els) |
Smush together the list of elements along an axis |
left.join(right, lsuffix='strL', rsuffix='strR') |
Combine diff df objects that have an index in common. left and right are df.s defined beforehand |
|
