Migrating from Python to C++
Using C++ for CGI scripts
Python is great for writing programs fast and I used Python 3 because there was a somewhat documented API for the Steem blockchain. However, it is only necessary for one script to run in the background and feed the API data to the database. I have always been partial to strong typing in languages. I am a big fan of Rust, but I don't really know all of the APIs and I fear they might be lacking. In compiled languages syntax errors are caught before you run tests on them and type-saftey prevents other non-syntax problems. Many problems that would throw a run-time exception in Python are caught at compile-time in C++.
Note: When I use three dots on one line in a row, it indicates there are more lines after or before the excerpt shown in the example.
Before I had python file for the configuration. This is because python is so simple, it looks so much like a config file.
Old steemfiles.py
root_dir = "/var/steemfiles/"
data_dir = root_dir + "data/"
bits_dir = root_dir + "bits/"
log_dir = root_dir + "log/"
The above is an except from the actual python code file. And essentially, converting it to an ini file is just getting rid of operations other than = and adding [all] to the top of the file. Now, all of the scripts are expecting these variables root_dir, data_dir, etc...
new steemfiles.ini
[all]
root_dir = "/var/steemfiles/"
data_dir = /var/steemfiles/data/
bits_dir = /var/steemfiles/bits/
log_dir = /var/steemfiles/log/
Similar, isn't it? Then we need python code to pull information out of this ini file. In C++ we will do the same to the same file.
Here is the code for loading the configuration of the file.
new steemfiles.py
import configparser
import os
import json
config = configparser.ConfigParser()
config.read(os.path.dirname(os.path.realpath(__file__)) + '/steemfiles.ini')
config = config['all']
Later on in the file we have these lines so we have same interface to all of the existing python code
# the data directory. Where user's files are stored
data_dir = config["data_dir"]
# Where template s and bits are stored
bits_dir = config["bits_dir"]
log_dir = config["log_dir"]
Now for boolean options I decided to validate things a bit better. After all, there is only two possible values, right?
try:
assert config['testnet'] == 'True' or config['testnet'] == 'False'
testnet = config['testnet'] == 'True'
except KeyError:
raise Exception("testnet not set in config file: Must be True or False")
Where's the C++?
I had to port Python into an INI file so that I can use this parser that works for ini file. Depending on the response I get from this post here I will decide if it is worth doing the next one. Comment, vote, resteem.
#include <vector>
#include <string>
#include <boost/program_options/config.hpp>
#include <boost/program_options/options_description.hpp>
#include <boost/program_options/parsers.hpp>
#include <boost/program_options/variables_map.hpp>
#include <fstream>
#include <iostream>
namespace po = boost::program_options;
using namespace std;
#include "steemfiles-config.hpp"
struct config_type steemfiles_setup(string program_name) {
// configuration as a C++ structure
config_type cfgs;
// description of configuration file
boost::program_options::options_description cfg_od("Options", 80, 40);
// configuration as a string to any map
po::variables_map cfgm;
try {
cfg_od.add_options()
("all.webserver_host", po::value< std::vector<std::string> >(), "localhost or steemfiles.com")
("all.root_dir", po::value< vector<string> >(), "root_dir")
// the data directory. Where user's files are stored
("all.data_dir", po::value< vector<string> >(), "")
// Where template s and bits are stored
("all.bits_dir", po::value< vector<string> >(), "")
("all.log_dir", po::value< vector<string> >(), "")
// much deleted...
("all.testnet", po::value< vector<string> >(), "");
Well, here is a cute way to parse not only configuration files but also command lines. We pass the type of the value if it is a numeric value. Boolean values are not handled well in this Boost C++ library so we make that a string and after we convert them into bool values.
Now just like in python I want these values to be stored into a global variable but instead of seperate variables without a type. I put all of the values in a struct. The struct gives everything a type.
static std::runtime_error webserver_host_not_in_cfg("webserver_host not specified in the config file!");
static std::runtime_error webserver_host_repeated_in_cfg("webserver_host specified more than once in the config file!");
if (cfgm.count("webserver_host") == 0) {
throw webserver_host_not_in_cfg;
} else if (cfgm.count("webserver_host") > 1) {
throw webserver_host_repeated_in_cfg;
} else {
cfgs.webserver_host = cfgm["webserver_host"].as< vector< string> >()[0];
It becomes tedious to write these again and again, so I right away I decided to make macros. After some refactoring, I settled on these macros:
#define check_option(opt) \
static std::runtime_error opt ## _repeated_in_cfg(#opt " specified more than once in the config file!");\
if (cfgm.count("all." #opt) == 0) {\
throw new std::runtime_error(string(#opt) + " not specified in the config file!");\
} else if (cfgm.count("all." #opt) > 1) {\
throw new std::runtime_error(string(#opt) + " specified more than once in the config file!");\
} else
#define check_option_string(opt) check_option(opt) {\
cfgs.opt = cfgm["all." #opt].as< vector< string> >()[0];\
}
#define check_option_int(opt) check_option(opt) {\
cfgs.opt = cfgm["all." #opt].as< int >();\
}
#define check_option_bool(opt) check_option(opt) {\
vector< string > vs = cfgm["all." #opt].as< vector< string > >();\
if (vs[0] == "False") {\
cfgs.opt = false;\
} else if (vs[0] == "True") {\
cfgs.opt = true;\
} else {\
throw new std::runtime_error(string(#opt) + " has an invalid value in the configuration file: It should be True or False not " + vs[0]);\
}\
}
After that, you can have options like these:
check_option_string(webserver_host)
check_option_string(root_dir)
check_option_string(data_dir)
// Where template s and bits are stored
check_option_string(bits_dir)
check_option_string(log_dir)
check_option_string(thumb_dir)
The header file is just a struct and declaration of this routine:
#include <string>
using namespace std;
struct config_type {
string webserver_host;
string root_dir;
// the data directory. Where user's files are stored
string data_dir;
// Where template s and bits are stored
string bits_dir;
string log_dir;
string thumb_dir;
//....
};
struct config_type steemfiles_setup(string program_name);
Finally
Well I hope that was interesting to some of you. I have just finished porting over the download.cgi script to C++. It is so much faster, now. It's still running on my first test-server at the moment.
Dash XjzjT4mr4f7T3E8G9jQQzozTgA2J1ehMkV
LTC LLXj1ZPQPaBA1LFtoU1Gkvu5ZrxYzeLGKt
BitcoinCash 1KVqnW7wZwn2cWbrXmSxsrzqYVC5Wj836u
Bitcoin 1Q1WX5gVPKxJKoQXF6pNNZmstWLR87ityw (too expensive to use for tips)
See my other recent article:
