01 numpy introduction

What is? NumPy（Numerical Python）

        NumPy yes Python The basic package of scientific computing in , It's a Python library , Provide multidimensional array objects , Various derived objects （ Such as mask array and matrix ）, And various methods for fast array operation API, Include mathematics 、 Logic 、 Shape operation 、 Sort 、 choice 、 Input and output 、 Discrete Fourier transform 、 Basic linear algebra , Basic statistical operations and stochastic simulation etc. .NumPy Fully support object-oriented Methods .

        NumPy The forerunner of Numeric The first is the Jim Hugunin Developed with other collaborators ,2005 year ,Travis Oliphant stay Numeric Another library of the same nature is combined in Numarray The characteristics of , And added other extensions and developed NumPy.NumPy Developed for open source and maintained by many collaborators .

        NumPy The history of can be traced back to 90 The mid - , Its predecessor is Numeric（ use C Language writing , It is mainly used to fetch C++ Application in ） and Numarray（ Used to process high-dimensional arrays , Flexible indexing 、 Data type transformation 、 Radio, etc ）,2005 Years of NumPy As a successor , Absorbed Numeric Rich in C API And Numarray High dimensional array processing power , Become Python Scientific computing is the foundation of ecosystem

        NumPy The core of the package is ndarray object . It encapsulates the python Of the same data type n Dimension group , In order to ensure its excellent performance , Many of these operations are performed after the code is compiled locally .

        NumPy It's a very fast math library , Mainly used for array calculation , contain ：

• A powerful N Dimensional array object ndarray
• Broadcast function
• Integrate C/C++/Fortran Code tools
• linear algebra 、 The Fourier transform 、 Random number generation and other functions

Why? NumPy So fast

1, The bottom is C, Part of it is Fortran

2, In addition to using C outside ,numpy Has its own set of instructions . It can improve the running speed .

3, stay R There is a thing called vectorization programming . Speed is directly based on embedded data structures 100 About times .

         Vectorization describes that there are no explicit loops in the code , Index, etc. . These are precompiled C In the code “ Behind the scenes ” Results of optimization . Without vectorization , Our code will be unreadable and inefficient for Troubled by the cycle . Vectorization code has many advantages , These include ：

• Vectorization code is more concise , Easier to read
• Fewer lines of code usually means fewer errors
• The code is closer to the standard mathematical notation （ Usually , It is easier to correctly code the mathematical structure ）
• Vectorization leads to more “Pythonic” Code .

         Generally speaking , stay NumPy in , All operations , Not just arithmetic , It is Logic , position , function etc. , In this implicit element by element way , That is, they broadcast .