Programming – StrayWarrior's Blog

A bug Caused by Implicit Construction of std::string

Recently I met a bug when I was using google protobuf to serialize and deserialize messages.I have a code segment like this:

class MyMessage; // A message class generated by protobuf</code>
void process() {
    MyMessage msg;
    std::string msg_buffer = some_other_func(); // receive binary format "string"
    msg.ParseFromString(msg_buffer.data());
}

class MyMessage; // A message class generated by protobuf</code>

void process() {

MyMessage msg;

std::string msg_buffer = some_other_func(); // receive binary format "string"

msg.ParseFromString(msg_buffer.data());

}

In the code, there is a weird usage of std::string::data() and MyMessage::ParseFromString() API. It’s a mistake due

Be careful with return-value optimization of C++ compilers

Side Effect of Return-value Optimization This “bug” is caused by return-value optimization and affects both GCC and Microsoft C/C++ Compiler. Assume that we have codes like following:

#include <cstdio>
class Complex {
public:
    Complex() : Complex(0.0, 0.0) {}
    Complex(double a, double b) : a(a), b(b) {}
    Complex(const Complex & x) : a(x.a), b(x.b) {
        dummy = 1.f;
    }
    Complex operator * (const Complex & x) {
        return Complex(a * x.a, b * x.b);
    }

    double a, b;
    double dummy;
};

int main() {
    Complex z1(-1, 2);
    Complex z2(3, 5);
    Complex z3 = z1 * z2;
    Complex z4 = z3;
    printf("%f %f %f\n", z3.a, z3.b, z3.dummy);
    printf("%f %f %f\n", z4.a, z4.b, z4.dummy);
    return 0;
}

#include <cstdio>

class Complex {

public:

Complex() : Complex(0.0, 0.0) {}

Complex(double a, double b) : a(a), b(b) {}

Complex(const Complex & x) : a(x.a), b(x.b) {

dummy = 1.f;

}

Complex operator * (const Complex & x) {

return Complex(a * x.a, b * x.b);

}

double a, b;

double dummy;

};

int main() {

Complex z1(-1, 2);

Complex z2(3, 5);

Complex z3 = z1 * z2;

Complex z4 = z3;

printf("%f %f %f\n", z3.a, z3.b, z3.dummy);

printf("%f %f %f\n", z4.a, z4.b, z4.dummy);

return 0;

}

Compile the code with g++ main.cpp -O0 -g -std=c++11. Run the program and the output result is

Compile and Install Basemap in Windows (Using MSVC)

Basemap is a Python package, which never releases new update after 2014. It’s easy to use in Linux or OS X but not easy to install on Windows. It doesn’t supply setup-package for Python 3.4 or 3.5 or newer version.

Link Error When Compiling and Installing Python Package in Windows

It’s usually easy to install a Python package in Windows with pip or other pre-built binaries. However, sometimes we must build the binary from source by ourselves, such as Basemap module, a part of matplotlib-toolkit.Then we may meet some error

Python 3使用Basemap导入ESRI信息的问题

Basemap使用自带的shapefile.py处理ESRI的shapefile，但是Basemap自带的shapefile版本过低(1.1.x)，其中存在一个bug导致使用Python 3.x版本时无法正确处理Windows-1252编码的二进制文件。典型错误：

File "D:\Python34\lib\site-packages\mpl_toolkits\basemap\shapefile.py", line 58, in u
    return v.decode('utf-8')
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xed in position 2: invalid continuation byte

File "D:\Python34\lib\site-packages\mpl_toolkits\basemap\shapefile.py", line 58, in u

return v.decode('utf-8')

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xed in position 2: invalid continuation byte

解决方法一：使用pip安装pyshp ( pip install pyshp )，或者手动下载最新版本的pyshp，并且将shapefile.py拷贝到Basemap的目录中( <python-install-prefix>\lib\site-packages\mpl_toolkits\basemap )。解决方法二：按照一的方式使用pip安装pyshp，并且修改Basemap的源代码，将__init__.py中shapefile模块的import语句略作修改，不再使用相对路径的导入方式并删除Basemap目录中的shapefile.py。修改之前为

from . import shapefile as shp
from .shapefile import Reader

1 2	from . import shapefile as shp from .shapefile import Reader

修改之后为

import shapefile as shp
from shapefile import Reader

1 2	import shapefile as shp from shapefile import Reader

使用方法二的一个好处是，使用pip更新pyshp后不需要再更新Basemap中的shapefile.py

Matlab求解变参数常微分方程组

以下代码在Matlab R2010a运行通过。待求解函数示例：

function dx = r1( t, x, k )
    dx = -k*x;
end

function dx = r1( t, x, k )

dx = -k*x;

end

k为可变参数。求解主程序：

sol = ode45(@r1, [0,10], 1, [], 1)
deval(sol, [2], 1)

1 2	sol = ode45(@r1, [0,10], 1, [], 1) deval(sol, [2], 1)

如上所示，在调用ODE solver时，格式为 solver(odefun, [t0 tfinal], y0, options, args...) 其中 args... 为可变参数，将被转发至待求解的方程中。 Matlab的自带帮助中并没有明确说明ODE solver支持这样的调用形式，不过这样的调用形式是符合Matlab函数定义的习惯的。上例中，ode45函数的最后一个参数(args…)为1，该参数被转发至待求解的方程中作为参数k。上述求解过程使用了 deval() 来返回t=2时的结果，此函数的第一个实参是sol，是一个structure，要让ODE solver返回该类型，则ODE solver必须采用函数句柄的方式输入待求解的函数，若使用inline函数将无法返回该structure而仅返回积分区间。关于该structure的具体定义可查看Matlab帮助。

已知平衡数据利用Matlab作三元相图

需准备：平衡线数据，连接线数据 1.作直角三角形相图

%limitline.csv 三列数据依次为平衡线上各点S,B,A的质量分数
%tieline.csv 六列数据依次为连接线上两点的S,B,A的质量分数
MLine = csvread('limitline.csv');
TLine = csvread('tieline.csv');
MLine = MLine ./100;
TLine = TLine ./100;
X = 0:0.01:1;
Y = interp1(MLine(:,1),MLine(:,3),X,'cubic');

plot([0,1],[1,0]);hold on;
plot(X,Y),axis([0 1 0 1]),axis equal,axis square

pointNumber = size(TLine);

for k = 1:pointNumber(1)
plot([TLine(k,1),TLine(k,4)],[TLine(k,3),TLine(k,6)])
end

%limitline.csv 三列数据依次为平衡线上各点S,B,A的质量分数

%tieline.csv 六列数据依次为连接线上两点的S,B,A的质量分数

MLine = csvread('limitline.csv');

TLine = csvread('tieline.csv');

MLine = MLine ./100;

TLine = TLine ./100;

X = 0:0.01:1;

Y = interp1(MLine(:,1),MLine(:,3),X,'cubic');

plot([0,1],[1,0]);hold on;

plot(X,Y),axis([0 1 0 1]),axis equal,axis square

pointNumber = size(TLine);

for k = 1:pointNumber(1)

plot([TLine(k,1),TLine(k,4)],[TLine(k,3),TLine(k,6)])

end

2.作等边三角形相图

%limitline.csv 三列数据依次为平衡线上各点S,B,A的质量分数
%tieline.csv 六列数据依次为连接线上两点的S,B,A的质量分数
MLine = csvread('limitline.csv');
TLine = csvread('tieline.csv');
MLine = MLine ./100;
TLine = TLine ./100;
X = 0:0.01:1;
Y = interp1(MLine(:,1)+0.5*MLine(:,3),MLine(:,3)*sqrt(3)/2,X,'cubic');

plot([0,0.5],[0,sqrt(3)/2]);hold on;
plot([1,0.5],[0,sqrt(3)/2]);hold on;
plot(X,Y),axis([0 1 0 1]),axis equal,axis square

pointNumber = size(TLine);

for k = 1:pointNumber(1)
plot([TLine(k,1)+0.5*TLine(k,3),TLine(k,4)+0.5*TLine(k,6)]，[TLine(k,3)*sqrt(3)/2,TLine(k,6)*sqrt(3)/2])
end