====== Useful python stuff ====== You will find on this page some **useful, but unsorted, python tips and tricks** that can't fit in a section of the main [[other:python:jyp_steps|JYP's recommended steps for learning python]] page ===== Extra tutorials ===== Only **when you have already read all the content of this page several times**, and you are looking for new ideas * [[https://medium.com/pythons-gurus/clean-code-in-python-good-vs-bad-practices-examples-2df344bddacc|Clean Code in Python: Good vs. Bad Practices Examples]] * [[https://peps.python.org/pep-0008/|PEP 8 – Style Guide for Python Code]] * [[https://realpython.com/python-pep8/|How to Write Beautiful Python Code With PEP 8]] * [[https://www.datacamp.com/tutorial/pep8-tutorial-python-code|PEP-8 Tutorial: Code Standards in Python]] * Some checkers/linters: [[https://docs.astral.sh/ruff/|ruff]], [[https://flake8.pycqa.org/en/stable/|flake8]] * [[https://medium.com/@yaduvanshineelam09/ultimate-python-cheat-sheet-practical-python-for-everyday-tasks-8a33abc0892f|Ultimate Python Cheat Sheet: Practical Python For Everyday Tasks]] * [[https://medium.com/pythoneers/16-hacks-that-will-take-your-python-skills-to-the-next-level-12e7a9b97421|16 Hacks That Will Take Your Python Skills to the Next Level]] * [[https://levelup.gitconnected.com/modular-coding-in-python-finally-solve-your-import-errors-af2fd172fcf7|Modular Coding in Python: Finally Solve your Import Errors]] (understanding and fixing ModuleNotFoundError and ImportError) * [[https://medium.com/@moraneus/understanding-multithreading-and-multiprocessing-in-python-1ed39bb078d5|Understanding Multithreading and Multiprocessing in Python]] ===== Reading/setting environments variables ===== >>> os.environ['TMPDIR'] '/data/jypmce/climafcache' >>> os.environ.get('SCRATCHDIR', '/data/jypmce/some_scratch_stuff') '/data/jypmce/some_scratch_stuff' >>> os.environ['temporary_env_var_for_THIS_script'] = 'some value' >>> os.environ['temporary_env_var_for_THIS_script'] 'some value' ===== Generating (aka raising) an error ===== This will stop the script, unless it is called in a function, and the code calling the function explicitely catches and deals with errors * raise RuntimeError('\n\nOMG! An error! :-(\nAborting script...') * [[https://docs.python.org/3/tutorial/errors.html|Errors and Exceptions tutorial]] * [[https://docs.python.org/3/library/exceptions.html|Built-in Exceptions reference]] ===== Using log files (aka logging) ===== It is always possible to display information messages using the ''print()'' command, but it is more efficient to use //logging// tools when you want to **display correctly a lot of information about a script progress ** * [[https://loguru.readthedocs.io/|Loguru]] is a library which aims to bring enjoyable logging in Python * See also [[https://betterstack.com/community/guides/logging/loguru/|A Complete Guide to Logging in Python with Loguru]] * More on [[https://betterstack.com/community/guides/logging/#python|logging with python]] * The default (but not easy to use) Python ''[[https://docs.python.org/3/library/logging.html|logging]]'' module ===== Stopping a script ===== A user can use ''CTRL-C'' or ''kill'' to stop a script, or ''CTRL-Z'' to suspend it temporarily (use ''fg'' to resume a suspended script). The code below can be used by the script itself to interrupt its execution, instead of raising an error sys.exit('Some optional message about why we are stopping') ===== Checking if a file/directory is writable by the current user ===== >>> os.access('/', os.W_OK) False >>> os.access('/home/jypmce/.bashrc', os.W_OK) True ===== Playing with strings ===== ==== String formatting ==== * Knowing how to display/print a string correctly is always useful for information and debugging purpose * There are lots of different ways to display strings === String formatting examples === You will find below some examples of //quick printing//, as well as using //old style formatting//, //formatted string literals (f-strings)// and the //String ''format()'' Method//. More details in the next section >>> # Basic (but quick and efficient) printing >>> year = 1984 >>> print(year) 1984 >>> print('[', year, 'is a famous book ]') [ 1984 is a famous book ] >>> # Old style formatting >>> print('[ %i is a famous book ]' % (year,)) [ 1984 is a famous book ] >>> print('[ %10i is a famous book ]' % (year,)) [ 1984 is a famous book ] >>> print('[ %-10i is a famous book ]' % (year,)) [ 1984 is a famous book ] >>> print('[ %010i is a famous book ]' % (year,)) [ 0000001984 is a famous book ] >>> # Formatted string literals (f-strings) >>> print(f'[ {year} is a famous book ]') [ 1984 is a famous book ] >>> print(f'[ {year=} is a famous book ]') [ year=1984 is a famous book ] >>> print(f'[ {year:10} is a famous book ]') [ 1984 is a famous book ] >>> print(f'[ {year:<10} is a famous book ]') [ 1984 is a famous book ] >>> print(f'[ {year:010} is a famous book ]') [ 0000001984 is a famous book ] >>> print(f'[ {year:10.2f} is a famous book (yes, {year}!) ]') [ 1984.00 is a famous book (yes, 1984!) ] >>> # The String format() Method >>> print('[ {} is a famous book ]'.format(year)) [ 1984 is a famous book ] >>> print('[ {:10} is a famous book ]'.format(year)) [ 1984 is a famous book ] >>> print('[ {:<10} is a famous book ]'.format(year)) [ 1984 is a famous book ] >>> print('[ {:010} is a famous book ]'.format(year)) [ 0000001984 is a famous book ] >>> print('[ {:10.2f} is a famous book (yes, {}!) ]'.format(year, year)) [ 1984.00 is a famous book (yes, 1984!) ] >>> print('[ {title:10.2f} is a famous book (yes, {title}!) ]'.format(title=year)) [ 1984.00 is a famous book (yes, 1984!) ] >>> print('[ {title:10.2e} is a famous book ]'.format(title=year)) [ 1.98e+03 is a famous book ] === String formatting references === * [[https://docs.python.org/3/tutorial/inputoutput.html#formatted-string-literals|Formatted String Literals]] (//f-strings//) * Available in Python >= 3.6 * [[https://docs.python.org/3/reference/lexical_analysis.html#f-strings|More documentation]] * [[https://docs.python.org/3/library/string.html#formatspec|Format Specification Mini-Language]] * See also the [[https://pyformat.info/|PyFormat site]] * [[https://docs.python.org/3/tutorial/inputoutput.html#the-string-format-method|The String format() Method]] * [[https://docs.python.org/3/library/string.html#formatspec|Format Specification Mini-Language]] * See also the [[https://pyformat.info/|PyFormat site]] * [[https://pyformat.info/|PyFormat site]]: string formatting using the //old style// and the //String ''format()'' method// * Hint: this can also be used as an **easy documentation for //f-strings// format**! * [[https://docs.python.org/3/tutorial/inputoutput.html#the-string-format-method|Old string formatting]] ==== Splitting (complex) strings ==== It's easy to split a string with multiple blank delimiters, or a specific delimiter, but it can be harder to deal with sub-strings >>> str_with_blanks = 'one two\t3\t\tFOUR' >>> str_with_blanks.split() ['one', 'two', '3', 'FOUR'] >>> str_with_simple_delimiters = '1,2,3.14, 4' >>> str_with_simple_delimiters.split(',') ['1', '2', '3.14', ' 4'] >>> complex_string='-o 1 --long "A string with accented chars: é è à ç"' >>> complex_string.split() ['-o', '1', '--long', '"A', 'string', 'with', 'accented', 'chars:', '\xc3\xa9', '\xc3\xa8', '\xc3\xa0', '\xc3\xa7"'] >>> import shlex >>> shlex.split(complex_string) ['-o', '1', '--long', 'A string with accented chars: \xc3\xa9 \xc3\xa8 \xc3\xa0 \xc3\xa7'] ==== Working with paths and filenames ==== If you are in a hurry, you can just use string functions to work with paths and file names. You will need some specific objects and functions to check if a file exists, and similar operations. Check the libraries listed below, that can automatically deal with Unix-type paths on Linux and MacOS computers, and Windows-type paths on Windows computers * [[https://docs.python.org/3/library/os.path.html|os.path]]: //common pathname manipulations// * Available since... a long time! Use this if you want to avoid backward compatibility problems * Some functions are directly in [[https://docs.python.org/3/library/os.html|os]] //Miscellaneous operating system interfaces//\\ e.g. [[https://docs.python.org/3/library/os.html#os.remove|os.remove]] and [[https://docs.python.org/3/library/os.html#os.rmdir|os.rmdir]] * [[https://docs.python.org/3/library/pathlib.html|pathlib]]: a **more recent** //object-oriented// way to deal with //filesystem paths// * Available since Python version 3.4 * [[https://docs.python.org/3/library/pathlib.html#correspondence-to-tools-in-the-os-module|Matching pathlib, and os or os.path functions]] * [[https://docs.python.org/3/library/shutil.html|shutil]]: High-level file operations, e.g copy/move a file or directory tree === Example: getting the full path of the Python executable used === Note: the actual python may be different from the default python! $ which python /usr/bin/python $ /home/share/unix_files/cdat/miniconda3_21-02/envs/cdatm_py3/bin/python >>> import sys, shutil >>> shutil.which('python') '/usr/bin/python' >>> sys.executable '/home/share/unix_files/cdat/miniconda3_21-02/envs/cdatm_py3/bin/python' === Example: getting the full path of a script === >>> import os >>> os.getcwd() '/home/jypmce/PMIP4' >>> os.path.exists('./argv_test.py') True >>> os.path.abspath('./argv_test.py') '/home/jypmce/PMIP4/argv_test.py' >>> os.path.exists('/home/jypmce/PMIP4/argv_test.py') True === Example: system independent paths with pathlib === Note: the following example was generated on a Linux server and uses a / character as a path separator >>> my_home = Path.home() >>> my_home PosixPath('/home/users/my_login') >>> my_conf = my_home / '.config' / 'evince' >>> my_conf PosixPath('/home/users/my_login/.config/evince') >>> my_conf.is_dir() True >>> my_conf.is_file() False >>> list(my_conf.glob('*')) [PosixPath('/home/users/my_login/.config/evince/evince_toolbar.xml'), PosixPath(' /home/users/my_login/.config/evince/accels')] >>> [ ff.name for ff in my_conf.glob('*') ] ['evince_toolbar.xml', 'accels'] === Example: getting the size(s) of all the files in a directory === $ cd /data/jypmce/TestDir $ ls -l total 72 -rw-r--r-- 1 jypmce ipsl 18147 Jun 25 2012 get_TS_cmip5.py -rw-r--r-- 1 jypmce ipsl 16152 Jun 21 2012 get_TS_cmip5.py~ -rw-r--r-- 1 jypmce ipsl 13954 Jul 3 2012 get_TS_cmip5_regular.py -rw-r--r-- 1 jypmce ipsl 16539 Jun 22 2012 get_TS_cmip5_regular.py~ >>> os.chdir('/data/jypmce/TestDir') >>> print(os.getcwd()) /data/jypmce/TestDir >>> files_list = os.listdir() >>> files_list ['get_TS_cmip5.py~', 'get_TS_cmip5_regular.py', 'get_TS_cmip5_regular.py~', 'get_TS_cmip5.py'] >>> files_sizes = list(map(os.path.getsize, files_list)) >>> files_sizes [16152, 13954, 16539, 18147] >>> sum(files_sizes) 64792 ==== Generating file names ==== === Name depending on the current date/time === >>> import time >>> plot_version = time.strftime('%Y%m%d_%H%M') >>> f_name = 'test_%s.nc' % (plot_version,) >>> f_name 'test_20210827_1334.nc' === Temporary file === >>> import tempfile, os >>> f_tmp = tempfile.NamedTemporaryFile(mode='w', suffix='.nc', delete=False) >>> f_tmp >>> f_tmp.name '/tmp/tmpi6uk9hre.nc' >>> f_tmp.close() >>> os.remove(f_tmp.name) ===== Using command-line arguments ===== ==== The extremely easy but non-flexible way: sys.argv ==== The name of a script, the number of arguments (including the name of the script), and the arguments (as strings) can be accessed through the ''sys.argv'' strings' list Simple ''argv_test.py'' test script: #!/usr/bin/env python import sys nb_args = len(sys.argv) print('Number of script arguments (including script name) =', nb_args) for idx, val in enumerate(sys.argv): print(idx, val) $ python argv_test.py Number of script arguments (including script name) = 1 0 argv_test.py $ python argv_test.py tas tas_tes.nc Number of script arguments (including script name) = 3 0 argv_test.py 1 tas 2 tas_tes.nc ==== The C-style way: getopt ==== Use [[https://docs.python.org/3/library/getopt.html|getopt]] (//C-style parser for command line options//) ==== The deprecated Python way: optparse ==== [[https://docs.python.org/3/library/optparse.html|optparse]] (//parser for command line options//) is **deprecated since Python version 3.2**! You should now use argparse (check [[https://docs.python.org/3/library/argparse.html#upgrading-optparse-code|Upgrading optparse code]] for converting from ''optparse'' to ''argparse'') ==== The current Python way: argparse ==== [[https://docs.python.org/3/library/argparse.html|argparse]] (//parser for command-line options, arguments and sub-commands//) is available since Python version 3.2 ===== Using ordered dictionaries ===== **Dictionary order is guaranteed to be insertion order**! Note that the [[https://docs.python.org/3/library/stdtypes.html#dict|usual Python dictionary]] also guarantees the order since version **3.6** Check the [[https://docs.python.org/3/library/collections.html#collections.OrderedDict|OrderedDict class]] (''from collections import OrderedDict'') and the [[https://realpython.com/python-ordereddict/|OrderedDict vs dict in Python: The Right Tool for the Job]] tutorial ===== Using sets ===== [[https://docs.python.org/3/tutorial/datastructures.html#sets|Python sets]] are **groups of unique elements**. They can be used to easily find all the unique elements of //something// and you can easily determine the **intersection**, **union** (and other similar operations) of sets. ===== Printing a readable version of long lists or dictionaries ===== The [[https://docs.python.org/3/library/pprint.html|pprint]] module can be used for //pretty printing// objects (lists, dictionaries, ...). It will wrap long lines in a meaningful way >>> import pprint >>> test_dic = {'AWI-ESM-1-1-LR_AWI':{'r1i1p1f1': {'grid': 'gn'}}, 'CESM2_NCAR':{'r1i1p1f1': {'grid': 'gn'}}, 'IPSL-CM6A-LR_IPSL':{'r1i1p1f1': {'grid': 'gr'}, 'r1i1p1f2': {'grid': 'gr'}, 'r1i1p1f3': {'grid': 'gr'}, 'r1i1p1f4': {'grid': 'gr'}}} >>> print(test_dic) {'AWI-ESM-1-1-LR_AWI': {'r1i1p1f1': {'grid': 'gn'}}, 'CESM2_NCAR': {'r1i1p1f1': {'grid': 'gn'}}, 'IPSL-CM6A-LR_IPSL': {'r1i1p1f1': {'grid': 'gr'}, 'r1i1p1f2': {'grid': 'gr'}, 'r1i1p1f3': {'grid': 'gr'}, 'r1i1p1f4': {'grid': 'gr'}}} >>> pprint.pprint(test_dic) {'AWI-ESM-1-1-LR_AWI': {'r1i1p1f1': {'grid': 'gn'}}, 'CESM2_NCAR': {'r1i1p1f1': {'grid': 'gn'}}, 'IPSL-CM6A-LR_IPSL': {'r1i1p1f1': {'grid': 'gr'}, 'r1i1p1f2': {'grid': 'gr'}, 'r1i1p1f3': {'grid': 'gr'}, 'r1i1p1f4': {'grid': 'gr'}}} >>> dir(test_dic) ['__class__', '__contains__', '__delattr__', [... lots of unreadable stuff removed...] 'setdefault', 'update', 'values'] >>> pprint.pprint(dir(test_dic)) ['__class__', '__contains__', [... lots of lines removed in this example ] 'setdefault', 'update', 'values'] ===== Storing objects and data in a file (shelve and friends) ===== The built-in [[other:python:jyp_steps#the_shelve_package|shelve]] module can be **easily** used for storing temporary/intermediate data More options: * Some [[other:python:jyp_steps#data_file_formats|non-NetCDF]] file formats * Working with [[other:python:jyp_steps#netcdf_filesusing_cdms2_xarray_and_netcdf4|NetCDF]] files ===== Using a configuration file ===== The built-in [[https://docs.python.org/3/library/configparser.html|configparser]] module can be easily used for reading (**and** writing!) text configuration files. Note: a configuration file is also a way to easily store and exchange text data ! ===== Working with global variables ===== There is a good chance you don't actually want/need a //global// variable. Be sure to use the ''global'' statement correctly if you want to avoid side-effects... * [[https://docs.python.org/3/faq/programming.html?highlight=global#why-am-i-getting-an-unboundlocalerror-when-the-variable-has-a-value|Using (and changing) a global variable inside a script or module]] * Simple module example\\ _myvar = 10 def set_myvar(new_val): # Note: need to explicitly define a global variable (of a module) # as 'global' BEFORE changing its value in a function! # Otherwise, the value will not be REdefined outside the function global _myvar _myvar = new_val def get_myvar(): return _myvar def myfunc(nb_repeat = 10): print(nb_repeat * _myvar) * [[https://docs.python.org/3/faq/programming.html?highlight=global#how-do-i-share-global-variables-across-modules|Sharing global variables across modules]] ===== Sorting ===== * When dealing with **numerical values**, you should use the [[https://numpy.org/doc/stable/reference/routines.sort.html|numpy sorting, searching, and counting routines]]! * [[https://docs.python.org/3/howto/sorting.html|Sorting HOW TO]] * Example: sorting the keys and the values of a dictionary, and then using the ''key'' parameter to sort the keys of a dictionary according to the value associated with the key * If we provide a ''key'' function, the ''sort'' function will sort the elements by the values returned by the function, instead of sorting by the initial values. The function used for generating the key below is very simple and we can use a //lambda// (i.e //in place//) function * >>> demo_dic = {'a':10, 'b':5, 'c':-1, 'd':0} >>> sorted(demo_dic.keys()) ['a', 'b', 'c', 'd'] >>> sorted(demo_dic.values()) [-1, 0, 5, 10] >>> sorted(demo_dic.keys(), key=lambda key_name:demo_dic[key_name]) ['c', 'd', 'b', 'a'] ===== Efficient looping with numpy, map, itertools and list comprehension ===== Big, nested, explicit ''for'' loops should be avoided at all cost, in order to reduce a script execution time! * **''numpy'' arrays** should be used when dealing with //numerical data// * **Masked arrays** can be used to deal with //special cases// and remove tests from loops * The built-in [[https://docs.python.org/3/library/functions.html?highlight=map#map|map]] function (and similar functions like [[https://docs.python.org/3/library/functions.html?highlight=zip#zip|zip]], [[https://docs.python.org/3/library/functions.html?highlight=filter#filter|filter]], ...) can be used to efficiently apply a function (possibly a //simple// [[https://docs.python.org/3/tutorial/controlflow.html#lambda-expressions|lambda]] function) to all the elements of a list * >>> my_ints = [1, 2, 3] >>> map(str, my_ints) ['1', '2', '3'] >>> map(lambda ii: str(10*ii + 5), my_ints) ['15', '25', '35'] * The [[https://docs.python.org/3/library/itertools.html|itertools]] module defines many more fancy iterators that can be used for efficient looping * Example: replacing nested loops with [[https://docs.python.org/3/library/itertools.html#itertools.product|product]] * >>> it.product('AB', '01') >>> list(it.product('AB', '01')) [('A', '0'), ('A', '1'), ('B', '0'), ('B', '1')] >>> for c1, c2 in it.product('AB', '01'): ... print(c1 + c2) ... A0 A1 B0 B1 >>> for c1, c2 in it.product(['A', 'B'], ['0', '1']): ... print(c1 + c2) ... A0 A1 B0 B1 >>> for c1, c2, c3 in it.product('AB', '01', '$!'): ... print(c1 + c2 + c3, end=', ') ... A0$, A0!, A1$, A1!, B0$, B0!, B1$, B1!, * The [[https://docs.python.org/3/tutorial/datastructures.html?highlight=comprehension#list-comprehensions|list comprehension]] (aka //implicit loops//) can also be used to generate lists from lists * Example: converting a list of integers to a list of strings\\ Note: in that case, you should rather use the ''map'' function detailed above * >>> my_ints = [1, 2, 3] >>> [ str(ii) for ii in my_ints ] ['1', '2', '3'] ===== numpy related stuff ===== ==== Using a numpy array to store arbitrary objects ==== The numpy arrays are usually used to store [[https://numpy.org/doc/stable/reference/arrays.scalars.html|scalars]] of the same type (see also the [[https://numpy.org/doc/stable/reference/arrays.dtypes.html|Data type objects (dtype)]]), very often numerical values. It is also possible to store **arbitrary** Python objects in an array, rather than using nested lists or dictionaries! >>> some_array = np.empty((2, 3), dtype=object) >>> some_array array([[None, None, None], [None, None, None]], dtype=object) >>> some_array.shape (2, 3) >>> print(some_array[-1, -1]) None >>> some_array[-1, 0] = filled_contour # e.g. save an existing cartopy filled contour object >>> some_array array([[None, None, None], [, None, None]], dtype=object) ==== Dealing with a variable number of indices ==== [[https://numpy.org/doc/stable/user/basics.indexing.html#dealing-with-variable-indices|Official reference]] >>> i10 = np.identity(10) >>> i10 array([[1., 0., 0., 0., 0., 0., 0., 0., 0., 0.], [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.], ... [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]]) >>> i10.shape (10, 10) >>> i10[3:7, 4:6] array([[0., 0.], [1., 0.], [0., 1.], [0., 0.]]) >>> s0 = slice(3, 7) >>> s1 = slice(4, 6) >>> i10[s0, s1] array([[0., 0.], [1., 0.], [0., 1.], [0., 0.]]) >>> my_slices = (s0, s1) >>> i10[my_slices] array([[0., 0.], [1., 0.], [0., 1.], [0., 0.]]) >>> my_fancy_slices = (s0, Ellipsis) >>> i10[my_fancy_slices] array([[0., 0., 0., 1., 0., 0., 0., 0., 0., 0.], [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.], [0., 0., 0., 0., 0., 1., 0., 0., 0., 0.], [0., 0., 0., 0., 0., 0., 1., 0., 0., 0.]]) >>> i10[my_fancy_slices].shape (4, 10) >>> # WARNING! DANGERRRR! NEVER forget that a VIEW is NOT A COPY >>> # and that you can change the content of the original array by mistake >>> my_view = i10[my_slices] >>> my_view[:, :] = -1 >>> my_view array([[-1., -1.], [-1., -1.], [-1., -1.], [-1., -1.]]) >>> i10 array([[ 1., 0., 0., 0., 0., 0., 0., 0., 0., 0.], [ 0., 1., 0., 0., 0., 0., 0., 0., 0., 0.], [ 0., 0., 1., 0., 0., 0., 0., 0., 0., 0.], [ 0., 0., 0., 1., -1., -1., 0., 0., 0., 0.], [ 0., 0., 0., 0., -1., -1., 0., 0., 0., 0.], [ 0., 0., 0., 0., -1., -1., 0., 0., 0., 0.], [ 0., 0., 0., 0., -1., -1., 1., 0., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0., 1., 0., 0.], [ 0., 0., 0., 0., 0., 0., 0., 0., 1., 0.], [ 0., 0., 0., 0., 0., 0., 0., 0., 0., 1.]]) ==== Finding and counting unique values ==== Use ''np.unique'', do **not** try to use histogram related functions! >>> vals = np.random.randint(2, 5, (10,)) * 0.5 # Get 10 discreet float values >>> vals array([1. , 2. , 1. , 2. , 2. , 1.5, 1. , 1.5, 2. , 1.5]) >>> np.unique(vals) array([1. , 1.5, 2. ]) >>> unique_vals, nb_unique = np.unique(vals, return_counts=True) >>> unique_vals array([1. , 1.5, 2. ]) >>> nb_unique array([3, 3, 4]) >>> sorted_vals = np.sort(vals) # Sorted copy, in order to check the result >>> sorted_vals array([1. , 1. , 1. , 1.5, 1.5, 1.5, 2. , 2. , 2. , 2. ]) ==== Applying a ufunc over all the elements of an array ==== There are all sorts of //ufuncs// (Universal Functions), and we will just use below ''add'' from the [[https://numpy.org/doc/stable/reference/ufuncs.html#math-operations|math operations]], applied on the arrays defined in [[#finding_and_counting_unique_values|Finding and counting unique values]] # Get the sum of all the elements of 'vals' >>> np.add.reduce(vals) 15.5 >>> np.add.reduce(sorted_vals) 15.5 >>> vals.sum() # The usual and easy way to do it 15.5 # Compute the sum of the elements of 'nb_unique' # AND keep (accumulate) the intermediate results >>> nb_unique array([3, 3, 4]) >>> np.add.accumulate(nb_unique) array([ 3, 6, 10]) # The accumulated values can be used as indices to separate the different groups of sorted values! >>> sorted_vals array([1. , 1. , 1. , 1.5, 1.5, 1.5, 2. , 2. , 2. , 2. ]) >>> sorted_vals[0:3] array([1., 1., 1.]) >>> sorted_vals[3:6] array([1.5, 1.5, 1.5]) >>> sorted_vals[6:10] array([2., 2., 2., 2.]) # Compute the sum of each equal-value group >>> sorted_vals[0:3].sum(), sorted_vals[3:6].sum(), sorted_vals[6:10].sum() (3.0, 4.5, 8.0) ==== Applying a ufunc over specified sections of an array ==== The [[https://numpy.org/doc/stable/reference/generated/numpy.ufunc.reduceat.html#numpy.ufunc.reduceat|reduceat]] function can be used to avoid explicit python loops, and improve the speed (but not the readability...) of a script. The example below //improves// what has been shown above # Define a list with the boundaries of the intervals we want to apply the 'add' function to # We need to add the beginning index (0), AND remove the last index # (reduceat will automatically go to the end of the input array >>> nb_unique array([3, 3, 4]) >>> slices_indices = [0] + list(np.add.accumulate(nb_unique)) >>> slices_indices.pop() # Remove last element 10 >>> slices_indices [0, 3, 6] # Compute the sums over the selected intervals with just one call >>> np.add.reduceat(np.sort(vals), slices_indices) array([3. , 4.5, 8. ]) ==== Exercise your brain with numpy ==== Have a look at [[https://github.com/rougier/numpy-100/blob/master/100_Numpy_exercises.ipynb|100 numpy exercises]] ===== matplotlib related stuff ===== ==== Working with time axes (and ticks) ==== If you have problems setting the limits of a time axis, choosing the ticks' locations, or specifying the style of the labels, you should check the: * [[https://matplotlib.org/stable/gallery/index.html#ticks|Ticks examples' gallery]] * [[https://matplotlib.org/stable/gallery/text_labels_and_annotations/date.html|Date tick labels example]] ===== Data representation ===== A few notes for a future section or page about about //data representation// (bits and bytes) on disk and in memory, vs //data format// FIXME Add parts (pages 28 to 37) of this [[http://www.lsce.ipsl.fr/Phocea/file.php?class=page&file=5/pythonCDAT_jyp_2sur2_070306.pdf|old tutorial]] to this section ==== Base notions ==== * **Never forget** that all the bits and pieces of information we use are coded in [[https://en.wikipedia.org/wiki/Binary_number#Counting_in_binary|base 2]] (''0''s and ''1''s ...), grouped in bytes! * Some things can be stored exactly (integers, characters, ...) * In other cases (**//real// numbers** that we work with all the time, compressed images/videos/music) we only store **//good enough approximation//** * 1 byte <=> 8 bits * ''REAL*4'' <=> 4 bytes <=> 32 bits * For easier written/displayed representation, 1 byte is usually split into 2 groups of 4 bits, and displayed using base 16 and [[https://en.wikipedia.org/wiki/Hexadecimal|hexadecimal representation]] (characters ''0'', ''1'', ..., ''A'', ''B'', ..., ''F'') * ''0000'' <=> ''0'',\\ ''0010'' <=> ''1'', ...,\\ ''1111'' <=> ''F'' * ''1101'' <=> ''D'' in hexadecimal <=> ''13'' in decimal (''**1** * 8 + **1** * 4 + **0** * 2 + **1** * 1'') * ''11111101'' in //base 2// <=> ''1111 1101'' <=> ''FD'' in //hexadecimal// <=> ''253'' (''15 * 16 + 13'') in //decimal// * Base conversion with Python * >>> hex(13) # Decimal to Hexadecimal conversion '0xd' >>> hex(253) '0xfd' >>> hex(256) '0x100' >>> int('0x100', 16) # Hexadecimal to Decimal conversion 256 >>> int('1111', 2) # Binary to Decimal conversion 15 >>> int('11111101', 2) # '11111101' <=> '1111 1101' <=> 'FD' <=> 15 * 16 + 13 = 253 253 >>> 013 # DANGER! Python considers an integer to be in OCTAL base if it starts with a 0 11 >>> int('13', 8) # 1*8 + 3 11 * More technical topics * [[https://en.wikipedia.org/wiki/Bit_numbering|Bit numbering]]: the art of ordering bits, everything about MSB (Most Significant Byte) and LSB (Least Significant Byte) * [[https://en.wikipedia.org/wiki/Endianness|Endianness]]: the art of ordering bytes ==== Numerical values ==== * Binary data representation of some numbers (only some common types are listed here): * Languages and packages **references** used below: * Python: [[https://numpy.org/doc/stable/reference/arrays.scalars.html#sized-aliases|NumPy Sized aliases]] * NetCDF: [[https://docs.unidata.ucar.edu/nug/current/md_types.html|Data Types]], [[https://docs.unidata.ucar.edu/netcdf-fortran/current/f90-variables.html#f90-language-types-corresponding-to-netcdf-external-data-types|Fortran related Data Types]], [[https://docs.unidata.ucar.edu/nug/current/_c_d_l.html#cdl_data_types|CDL Data Types]] * Fortran: Intel Fortran Compiler [[https://www.intel.com/content/www/us/en/docs/fortran-compiler/developer-guide-reference/2023-1/intrinsic-data-types.html|Intrinsic Data Types]] * [[https://en.wikipedia.org/wiki/Integer_(computer_science)|Integers]] * Range: * 4-byte //signed// integers: ''−2,147,483,648'' to ''2,147,483,647'' * Python: ''numpy.int32'' * NetCDF: ''int'', ''NC_INT'' or ''NC_LONG'', ''NF90_INT'' * Fortran: ''INTEGER*4'' * 8-byte //signed// integers: ''−9,223,372,036,854,775,808'' to ''9,223,372,036,854,775,807'' * Python: ''numpy.int64'' * NetCDF: ''int64'', ''NC_INT64'' * Fortran: ''INTEGER*8'' * Tech note: signed integers use [[https://en.wikipedia.org/wiki/Two%27s_complement|two's complement]] for coding negative integers * [[https://en.wikipedia.org/wiki/IEEE_754|Floating point numbers]] (//IEEE 754// standard aka //IEEE Standard for Binary Floating-Point for Arithmetic//) * Range: * 4-byte float: ''~8 significant digits * 10E±38'' * Python: ''numpy.float32'' * NetCDF: ''float'', ''NC-FLOAT'', ''NF90_FLOAT'' * Fortran:''REAL*4'' * See also [[https://en.wikipedia.org/wiki/Single-precision_floating-point_format|Single-precision floating-point format]] * 8-byte float: ''~15 significant digits * 10E±308'' * Python: ''numpy.float64'' * NetCDF: ''double'', ''NC_DOUBLE'', ''NF90_DOUBLE'' * Fortran: ''REAL*8'' * **Special values**: * [[https://en.wikipedia.org/wiki/NaN|NaN]]: //Not a Number// * Python: ''numpy.nan'' * Infinity * Python: ''-numpy.inf'' and ''numpy.inf'' * Note: it is cleaner to use masks (and [[https://numpy.org/doc/stable/reference/maskedarray.generic.html|Numpy masked arrays]]) rather than ''NaN''s, when you have to deal with missing values ! * The RISKS of working with (the wrong) floats: * [[https://en.wikipedia.org/wiki/Round-off_error|Round-off error]] * [[https://en.wikipedia.org/wiki/Catastrophic_cancellation|Catastrophic cancellation]] * [[https://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html|What Every Computer Scientist Should Know About Floating-Point Arithmetic]] * A rather technical example: we //play// with a numpy 4-byte integer scalar * >>> one_int32 = np.int32(1) >>> one_int32 1 >>> type(one_int32) >>> one_int32.dtype dtype('int32') >>> one_int32.shape # A numpy SCALAR, is an ARRAY WITH NO SHAPE ! () >>> one_int32[0] Traceback (most recent call last): File "", line 1, in IndexError: invalid index to scalar variable. >>> one_int32[()] # Note how to access the single element, when there is NO SHAPE 1 >>> one_int32.ndim # NO SHAPE means no dimensions, but there is ONE element 0 >>> one_int32.size 1 >>> one_int32.nbytes # The element requires 4 bytes of storage 4 >>> hex(one_int32) # We can print the hexadecimal representation for INTEGERS scalars and arrays '0x1' >>> hex(one_int32 * 15) '0xf' >>> hex(one_int32 * 16) '0x10' # 'Serialize' the data (i.e. change the data to a series of bytes) # Note: the serialized data seems to be printed in the reverse order of 'hex(one_int32)' >>> one_int32_serialized = one_int32.tobytes() >>> type(one_int32_serialized) >>> len(one_int32_serialized) 4 >>> one_int32_serialized b'\x01\x00\x00\x00' >>> one_int32_serialized.hex(' ') # Another way to print the hexadecimal values '01 00 00 00' # Use the following in the unlikely case where you need to change the endianness (bytes ordering) >>> one_int32_reversed_endian = one_int32.byteswap() >>> one_int32_reversed_endian # Same bytes in a different order represent a different number (of course) 16777216 >>> hex(one_int32_reversed_endian) # Compare to the output of hex(one_int32) above '0x1000000' >>> one_int32_reversed_endian.tobytes() b'\x00\x00\x00\x01' * Another technical example: we use an array of 2 integers\\ When using ''byteswap()'', notice how bytes are swapped by groups of 4 bytes, because int32 use 4 bytes * >>> array_example = np.asarray((3, 17), dtype=np.int32) >>> array_example array([ 3, 17], dtype=int32) >>> array_example.shape, array_example.ndim, array_example.size, array_example.nbytes ((2,), 1, 2, 8) >>> array_example.tobytes().hex(' ', 4) '03000000 11000000' >>> array_example.byteswap().tobytes().hex(' ', 4) '00000003 00000011' * Manipulating binary data with [[https://docs.python.org/3/library/stdtypes.html#binary-sequence-types-bytes-bytearray-memoryview|bytes, bytearray, memoryview]] * Array addressing * [[https://www.geeksforgeeks.org/calculation-of-address-of-element-of-1-d-2-d-and-3-d-using-row-major-and-column-major-order/|Calculation of address of element of 1-D, 2-D, and 3-D using row-major and column-major order]] * In other words: //using indices to go from 1-D to n-Dimnensions data// * The [[https://en.wikipedia.org/wiki/Array_(data_structure)|array]] structure * python/C vs Fortran... * disk and ram usage: how to check the usage (available ram and disk), best practice on multi-user systems (how much allowed?) * ''du'', ''df'', ''cat /proc/meminfo'', ''top'' * understanding and reverse-engineering //binary// format * ''od'', ''strings'' * binary vs text format: ascii, utf, raw * text related functions in python: ''str'', ''int'', ''float'', ''ord'', ... * lists conversion with ''map'' and ''join'' * Misc : ''md5sum'' ==== Strings ==== * Encoding, [[https://en.wikipedia.org/wiki/ASCII|ASCII]], [[https://en.wikipedia.org/wiki/Unicode|unicode]], [[https://en.wikipedia.org/wiki/UTF-8|UTF-8]], ... * Getting the binary representation of a string * >>> test_string = 'A B 0 1 à µ' >>> type(test_string) >>> len(test_string) 11 >>> test_string_bin = test_string.encode('utf-8') >>> test_string_bin b'A B 0 1 \xc3\xa0 \xc2\xb5' >>> type(test_string_bin) >>> len(test_string_bin) 13 >>> test_string_bin.hex('-') '41-20-42-20-30-20-31-20-c3-a0-20-c2-b5' /* ===== Tip template ===== Some code */ /* standard page footer */ \\ \\ \\ ---- [ [[pmip3:|PMIP3 Wiki Home]] ] - [ [[pmip3:wiki_help|Help!]] ] - [ [[wiki:syntax|Wiki syntax]] ]