Related
I have a class assignment that asks us to write a C++ program that tracks spending, allows modification of the records, and returns "satisfaction" numbers about individual expenses (i.e. how good the user felt about spending that money). Our instructor has indicated that he'd like us to use sqlite3 in this program. He's given us a sample program that builds a table in sqlite3 and inputs predetermined values for the columns. This program runs just fine without issue.
What I am trying to do is modify the program to accept user inputs and store them in the sqlite3 database. This is the code I have thus far:
int main()
{
string salesDesc;
int price;
int satisf;
sqlite3 *db;
char *szErrMsg = 0;
cout << "Description of Expense: ";
cin >> saleDesc;
cout << endl;
cout << "Price: ";
cin >> price;
cout << endl;
cout << "Your Satisfaction: ";
cin >> satisf;
cout << endl;
// open database
int rc = sqlite3_open("spending_track.sqlite", &db);
if (rc)
{
cout << "Cannot open database\n";
}
else
{
cout << "Database opened successfully\n";
}
const char *pSQL[6];
pSQL[0] = "CREATE TABLE IF NOT EXISTS expenses(id INTEGER PRIMARY KEY "
"AUTOINCREMENT NOT NULL, logged TIMESTAMP DEFAULT "
"CURRENT_TIMESTAMP NOT NULL, desc VARCHAR(40), price INT,"
"satisfaction INT)";
pSQL[1] = "INSERT INTO expenses('" + string(saleDesc) + "'," price "," satisf ")";
pSQL[2] = "SELECT * FROM expenses";
pSQL[3] = "SELECT sum(satisf) FROM expenses";
// blablabla the rest of the program
When I try to compile this, I receive the following error:
error: cannot convert 'std::_cxx11::basic_string' to 'const char*' in assignment
pSQL[1] = "INSERT INTO expenses('" + string(saleDesc) + "'," price "," satisf ")";
If I change string(saleDesc) to saleDesc, I get the same error.
If I change string saleDesc; to char* saleDesc;, I receive the following error:
error: invalid operands of types 'const char[23]' and 'char*' to binary 'operator+'
pSQL[1] = "INSERT INTO expenses('" + string(saleDesc) + "'," price "," satisf ")";
I'm not sure what else to try to get this to work. I have also heard that it's a bad idea to allow users to directly input to sqlite3 tables. What would be a more "proper" way to do this?
Since this is just a class assignment, I doubt that you are going to have to worry about SQL injection attacks, so I wouldn't bother trying to sanitize your input.
Your other issue is you are confusing char*s and std::strings. The sqlite API requires you to pass it char*s so it can be used from C code, however that doesn't mean you need to use them. std::string is a wrapper for the char array, which you can get with the c_str() method. I don't think you really need to put the SQL statements in an array at the end. How about something like this:
std::string addTable = "CREATE TABLE IF NOT EXISTS expenses(id INTEGER PRIMARY KEY "
"AUTOINCREMENT NOT NULL, logged TIMESTAMP DEFAULT "
"CURRENT_TIMESTAMP NOT NULL, desc VARCHAR(40), price INT,"
"satisfaction INT)";
std::string insertExpense = "INSERT INTO expenses('" + saleDesc + "'," + std::to_string(price) "," + std::to_string(satisf) + ")";
std::string selectAllExpenses = "SELECT * FROM expenses";
Then when you want to pass it to the sqlite API, you could use c_str()
sqlite3_exec(db, addTable.c_str(), ...
Thanks everyone for the responses. I spent about an hour and a half with my professor yesterday going over this, and this actually stumped him. I eventually found a way to make this work with the array, but I want to stress that the solution I came up with is pretty much only good for this assignment. For anyone reading this with a similar problem, this method is not only messy, but also allows for SQL injection which should be avoided.
The problem, as many here have mentioned in comments, was that I was trying to stick a string into a char* array. The workaround we came up with was to add the SQL commands with the variables expanded in them directly to a string variable, like so:
string insertExpense = "INSERT INTO expenses(desc, price, satisf) VALUES ('" + saleDesc + "', "
""+ to_string(price) + ", " + to_string(satisf) + ")";
We then made that variable a c_str and assigned it to a char* variable, like so:
const char *line1 = insertExpense.c_str();
We then simply assigned this char* variable directly to the correct position in the array, like so:
const char *pSQL[6];
pSQL[0] = "CREATE TABLE IF NOT EXISTS expenses(id INTEGER PRIMARY KEY "
"AUTOINCREMENT NOT NULL, logged TIMESTAMP DEFAULT "
"CURRENT_TIMESTAMP NOT NULL, desc VARCHAR(40), price REAL,"
"satisf INT)";
pSQL[1] = line1;
pSQL[2] = "SELECT * FROM expenses";
pSQL[3] = "SELECT sum(satisf) FROM expenses";
This method correctly makes the SQL table and populates it with the correct statements as stored in their respective variables. I want to stress again that this method is both very messy and dangerous, and for anyone with a similar issue, it is probably a much better idea to use prepared statements, as others in the comments have already mentioned. Thank you everyone!
I have a program which reads information about 3D meshes in from a text file, stores the data as a mesh,performs some post-processing on the mesh, then collects information like volume, center of mass, etc.
The program has two threads:
Main thread starts the second thread, reads the file, does the processing, then waits for the second thread. As a part of the processing, it puts information about the meshes its just read onto a queue.
Second thread connects to SQL Server using Poco ODBC, puts some initial information about the file its reading into a database table, then gets information off the queue and assembles a potentially lengthy insert command. When it is done doing so, it submits the command, performs a final update command regarding the results of operations performed, then lets the main thread know it's done, and terminates the 2nd thread.
Everything works, right up until the moment it submits the large insert command. It throws an exception, and i can't figure out why.
Here i will give a simplistic outline of the code that is executing. Assume variables exist and are initialized.
The poco commands i run are:
using namespace Poco;
using namespace Poco::Data::Keywords;
using namespace Poco::Data;
ODBC::Connector::registerConnector();
Session session(SessionFactory::instance().create("ODBC", "Driver={SQL Server};Server=<hostname>;Database=<DB name>;Uid=<username>;Pwd=<password>;"));
session << "INSERT INTO TableName1 (SourceFileName,UserName) VALUES (?,?)",use(_filename),use(username),now;
session << "SELECT SCOPE_IDENTITY()", into(runID),now; //This always runs and returns 0.
string queryString = "Insert into TableName2 (Field1, field2, field3, field4, ...) "+
"VALUES (val1, val2, val3, val4, ...)"+
",(valA, valB, valC, valD, ...),..."
session << queryString,now;
Statement update(session);
update << "UPDATE TableName1 SET Field2 = ?, Field3 = ?, Field4 = ? WHERE Field1 = ?", use(data2), use(data3), use(data3), use(identity);
update.execute();
ODBC::Connector::unregisterConnector();
<send signal to main thread indicating being done.>
I'm trying to figure out a few key things.
How can I tell what state the Session is in?
Is there a way to ask Poco what went wrong and have it print an error message?
Are there any special things I need to set up to be able to specify a big insert statement all together in text like I am? I have tried it using ? placeholders, or executing individual statements, but it always gives me an exception on that part.
Is there a way to have statements execute under the same connection for sure? Normally I would do my INSERT INTO TableName1(...)VALUES(...) SELECT SCOPE_IDENTITY() all as a single operation. I've tried my commands in SQL Server Management Studio and it works properly. Right now, it is always returning a 0 aka NULL, like the statements run in separate connections.
More information:
String query = "INSERT INTO TableName1 (SourceFileName,UserName) VALUES ('"+ __file + "','" + __username + "')";
Statement insertStmt = (session << query);
try{insertStmt.execute(true);}
catch(Poco::Exception exc)
{
cout << exc.displayText() << endl;
}
try{session << "SELECT SCOPE_IDENTITY() as SCOPE_IDENTITY", into(runID), now;
cout << "Run ID: " << runID << endl;}
catch(Poco::Exception exc)
{
cout << exc.displayText() << endl;
}
I greatly appreciate your help or any suggestions on how I can improve this question.
1.:
There are various query members in the Session class - isConnected(), canTransact(), isTransaction() ... (if that is what you are looking for; if not, see the next answer)
1. and 2.:
Wrap your statement into try/catch block:
#include "Poco/Data/ODBC/ODBCException.h"
//...
try
{
session << "INSERT INTO TableName1 (SourceFileName, UserName) VALUES (?, ?) ",use(_filename),use(username),now;
}
catch(Poco::Data::ODBC::ConnectionException& ce){ std::cout << ce.toString() << std::endl; }
catch(Poco::Data::ODBC::StatementException& se){ std::cout << se.toString() << std::endl; }
3.:
I don't think the problem is too large statement. There is a configurable internal setting limiting the string size to 1K, but this applies to value strings, not the whole SQL statement.
If you still think that is the problem, you can increase the max field size, e.g.:
std::size_t fieldSize = 4096; //4K
session.impl()->setProperty("maxFieldSize", Poco::Any(fieldSize));
4.:
Poco::Data::ODBC does not parse or analyze the SQL statement in any way; so from that standpoint, whatever works with your ODBC driver will be fine.
As part of a bigger application I am working on a class for reading input from a text file for use in the initialization of the program. Now I am myself fairly new to programming, and I only started to learn C++ in December, so I would be very grateful for some hints and ideas on how to get started! I apologise in advance for a rather long wall of text.
The text file format is "keyword-driven" in the following way:
There are a rather small number of main/section keywords (currently 8) that need to be written in a given order. Some of them are optional, but if they are included they should adhere to the given ordering.
Example:
Suppose there are 3 potential keywords ordered like as follows:
"KEY1" (required)
"KEY2" (optional)
"KEY3" (required)
If the input file only includes the required ones, the ordering should be:
"KEY1"
"KEY3"
Otherwise it should be:
"KEY1"
"KEY2"
"KEY3"
If all the required keywords are present, and the total ordering is ok, the program should proceed by reading each section in the sequence given by the ordering.
Each section will include a (possibly large) amount of subkeywords, some of which are optional and some of which are not, but here the order does NOT matter.
Lines starting with characters '*' or '--' signify commented lines, and they should be ignored (as well as empty lines).
A line containing a keyword should (preferably) include nothing else than the keyword. At the very least, the keyword must be the first word appearing there.
I have already implemented parts of the framework, but I feel my approach so far has been rather ad-hoc. Currently I have manually created one method per section/main keyword , and the first task of the program is to scan the file for to locate these keywords and pass the necessary information on to the methods.
I first scan through the file using an std::ifstream object, removing empty and/or commented lines and storing the remaining lines in an object of type std::vector<std::string>.
Do you think this is an ok approach?
Moreover, I store the indices where each of the keywords start and stop (in two integer arrays) in this vector. This is the input to the above-mentioned methods, and it would look something like this:
bool readMAINKEY(int start, int stop);
Now I have already done this, and even though I do not find it very elegant, I guess I can keep it for the time being.
However, I feel that I need a better approach for handling the reading inside of each section, and my main issue is how should I store the keywords here? Should they be stored as arrays within a local namespace in the input class or maybe as static variables in the class? Or should they be defined locally inside relevant functions? Should I use enums? The questions are many!
Now I've started by defining the sub-keywords locally inside each readMAINKEY() method, but I found this to be less than optimal. Ideally I want to reuse as much code as possible inside each of these methods, calling upon a common readSECTION() method, and my current approach seems to lead to much code duplication and potential for error in programming. I guess the smartest thing to do would simply be to remove all the (currently 8) different readMAINKEY() methods, and use the same function for handling all kinds of keywords. There is also the possibility for having sub-sub-keywords etc. as well (i.e. a more general nested approach), so I think maybe this is the way to go, but I am unsure on how it would be best to implement it?
Once I've processed a keyword at the "bottom level", the program will expect a particular format of the following lines depending on the actual keyword. In principle each keyword will be handled differently, but here there is also potential for some code reuse by defining different "types" of keywords depending on what the program expects to do after triggering the reading of it. Common task include e.g. parsing an integer or a double array, but in principle it could be anything!
If a keyword for some reason cannot be correctly processed, the program should attempt as far as possible to use default values instead of terminating the program (if reasonable), but an error message should be written to a logfile. For optional keywords, default values will of course also be used.
In order to summarise, therefore, my main questions are the following:
1. Do you think think my approach of storing the relevant lines in a std::vector<std::string> to be reasonable?
This will of course require me to do a lot of "indexing work" to keep track of where in the vector the different keywords are located. Or should I work more "directly" with the original std::ifstream object? Or something else?
2. Given such a vector storing the lines of the text file, how I can I best go about detecting the keywords and start reading the information following them?
Here I will need to take account of possible ordering and whether a keyword is required or not. Also, I need to check if the lines following each "bottom level" keyword is in the format expected in each case.
One idea I've had is to store the keywords in different containers depending on whether they are optional or not (or maybe use object(s) of type std::map<std::string,bool>), and then remove them from the container(s) if correctly processed, but I am not sure exactly how I should go about it..
I guess there is really a thousand different ways one could answer these questions, but I would be grateful if someone more experienced could share some ideas on how to proceed. Is there e.g. a "standard" way of doing such things? Of course, a lot of details will also depend on the concrete application, but I think the general format indicated here can be used in a lot of different applications without a lot of tinkering if programmed in a good way!
UPDATE
Ok, so let my try to be more concrete. My current application is supposed to be a reservoir simulator, so as part of the input I need information about the grid/mesh, about rock and fluid properties, about wells/boundary conditions throughout the simulation and so on. At the moment I've been thinking about using (almost) the same set-up as the commercial Eclipse simulator when it comes to input, for details see
http://petrofaq.org/wiki/Eclipse_Input_Data.
However, I will probably change things a bit, so nothing is set in stone. Also, I am interested in making a more general "KeywordReader" class that with slight modifications can be adapted for use in other applications as well, at least it can be done in a reasonable amount of time.
As an example, I can post the current code that does the initial scan of the text file and locates the positions of the main keywords. As I said, I don't really like my solution very much, but it seems to work for what it needs to do.
At the top of the .cpp file I have the following namespace:
//Keywords used for reading input:
namespace KEYWORDS{
/*
* Main keywords and corresponding boolean values to signify whether or not they are required as input.
*/
enum MKEY{RUNSPEC = 0, GRID = 1, EDIT = 2, PROPS = 3, REGIONS = 4, SOLUTION = 5, SUMMARY =6, SCHEDULE = 7};
std::string mainKeywords[] = {std::string("RUNSPEC"), std::string("GRID"), std::string("EDIT"), std::string("PROPS"),
std::string("REGIONS"), std::string("SOLUTION"), std::string("SUMMARY"), std::string("SCHEDULE")};
bool required[] = {true,true,false,true,false,true,false,true};
const int n_key = 8;
}//end KEYWORDS namespace
Then further down I have the following function. I am not sure how understandable it is though..
bool InputReader::scanForMainKeywords(){
logfile << "Opening file.." << std::endl;
std::ifstream infile(filename);
//Test if file was opened. If not, write error message:
if(!infile.is_open()){
logfile << "ERROR: Could not open file! Unable to proceed!" << std::endl;
std::cout << "ERROR: Could not open file! Unable to proceed!" << std::endl;
return false;
}
else{
logfile << "Scanning for main keywords..." << std::endl;
int nkey = KEYWORDS::n_key;
//Initially no keywords have been found:
startIndex = std::vector<int>(nkey, -1);
stopIndex = std::vector<int>(nkey, -1);
//Variable used to control that the keywords are written in the correct order:
int foundIndex = -1;
//STATISTICS:
int lineCount = 0;//number of non-comment lines in text file
int commentCount = 0;//number of commented lines in text file
int emptyCount = 0;//number of empty lines in text file
//Create lines vector:
lines = std::vector<std::string>();
//Remove comments and empty lines from text file and store the result in the variable file_lines:
std::string str;
while(std::getline(infile,str)){
if(str.size()>=1 && str.at(0)=='*'){
commentCount++;
}
else if(str.size()>=2 && str.at(0)=='-' && str.at(1)=='-'){
commentCount++;
}
else if(str.size()==0){
emptyCount++;
}
else{
//Found a non-empty, non-comment line.
lines.push_back(str);//store in std::vector
//Start by checking if the first word of the line is one of the main keywords. If so, store the location of the keyword:
std::string fw = IO::getFirstWord(str);
for(int i=0;i<nkey;i++){
if(fw.compare(KEYWORDS::mainKeywords[i])==0){
if(i > foundIndex){
//Found a valid keyword!
foundIndex = i;
startIndex[i] = lineCount;//store where the keyword was found!
//logfile << "Keyword " << fw << " found at line " << lineCount << " in lines array!" << std::endl;
//std::cout << "Keyword " << fw << " found at line " << lineCount << " in lines array!" << std::endl;
break;//fw cannot equal several different keywords at the same time!
}
else{
//we have found a keyword, but in the wrong order... Terminate program:
std::cout << "ERROR: Keywords have been entered in the wrong order or been repeated! Cannot continue initialisation!" << std::endl;
logfile << "ERROR: Keywords have been entered in the wrong order or been repeated! Cannot continue initialisation!" << std::endl;
return false;
}
}
}//end for loop
lineCount++;
}//end else (found non-comment, non-empty line)
}//end while (reading ifstream)
logfile << "\n";
logfile << "FILE STATISTICS:" << std::endl;
logfile << "Number of commented lines: " << commentCount << std::endl;
logfile << "Number of non-commented lines: " << lineCount << std::endl;
logfile << "Number of empty lines: " << emptyCount << std::endl;
logfile << "\n";
/*
Print lines vector to screen:
for(int i=0;i<lines.size();i++){
std:: cout << "Line nr. " << i << " : " << lines[i] << std::endl;
}*/
/*
* So far, no keywords have been entered in the wrong order, but have all the necessary ones been found?
* Otherwise return false.
*/
for(int i=0;i<nkey;i++){
if(KEYWORDS::required[i] && startIndex[i] == -1){
logfile << "ERROR: Incorrect input of required keywords! At least " << KEYWORDS::mainKeywords[i] << " is missing!" << std::endl;;
logfile << "Cannot proceed with initialisation!" << std::endl;
std::cout << "ERROR: Incorrect input of required keywords! At least " << KEYWORDS::mainKeywords[i] << " is missing!" << std::endl;
std::cout << "Cannot proceed with initialisation!" << std::endl;
return false;
}
}
//If everything is in order, we also initialise the stopIndex array correctly:
int counter = 0;
//Find first existing keyword:
while(counter < nkey && startIndex[counter] == -1){
//Keyword doesn't exist. Leave stopindex at -1!
counter++;
}
//Store stop index of each keyword:
while(counter<nkey){
int offset = 1;
//Find next existing keyword:
while(counter+offset < nkey && startIndex[counter+offset] == -1){
offset++;
}
if(counter+offset < nkey){
stopIndex[counter] = startIndex[counter+offset]-1;
}
else{
//reached the end of array!
stopIndex[counter] = lines.size()-1;
}
counter += offset;
}//end while
/*
//Print out start/stop-index arrays to screen:
for(int i=0;i<nkey;i++){
std::cout << "Start index of " << KEYWORDS::mainKeywords[i] << " is : " << startIndex[i] << std::endl;
std::cout << "Stop index of " << KEYWORDS::mainKeywords[i] << " is : " << stopIndex[i] << std::endl;
}
*/
return true;
}//end else (file opened properly)
}//end scanForMainKeywords()
You say your purpose is to read initialization data from a text file.
Seems you need to parse (syntax analyze) this file and store the data under the right keys.
If the syntax is fixed and each construction starts with a keyword, you could write a recursive descent (LL1) parser creating a tree (each node is a stl vector of sub-branches) to store your data.
If the syntax is free, you might pick JSON or XML and use an existing parsing library.
cgicc can process form elements quite well , but how can i know whether the data is generated from get_method or post_method?
the piece of code i used:
cout << "Content-type:text/html\r\n\r\n";
try {
Cgicc cgi;
const_form_iterator iter;
for(iter = cgi.getElements().begin();
iter != cgi.getElements().end();
++iter){
cout <<
"<table><tr>" <<
"<td>" << iter->getName() << "</td>" <<
"<td>" << iter->getValue() << "</td>" <<
"</tr></table>" << endl;
}
}catch(exception& e) {
cout << e.what() << endl;
}
update:
i find this from the cgicc official page: "Parses both GET and POST form data transparently." (http://www.gnu.org/software/cgicc/)
it seems that cgicc don't want to separate get and post by design?
You can find the HTTP method (ì.e. GET, POST, etc...) of a request using cgicc::CgiEnvironment::getRequestMethod
I think the only way to solve it is by checking whether there is a variable name in the GET method query string that's the same name of the one in the POST method. This means that the variable name must be mentioned ONLY ONCE either of the two methods. In other words, if you combine the variables of the GET method with the variables of the POST in a single set, the variable name must be mentioned once in this set.
I'm building a graph generator using Boost Graph and Program Options. There are, for example, two types of components C and W, each with 1 source, 1 sink and some additional parameters to specify topology in between. I'd like to be able to stitch them together in the sequence provided by the order of the command line arguments.
For example:
./bin/make_graph -c4,5,1 -w3,3 -c3,1,2
Should create a graph resembling the following:
C -- W -- C
But:
./bin/make_graph -c4,5,1 -c3,1,2 -w3,3
Should create a graph resembling the following:
C -- C -- W
Using boost::program_options, I was unable to determine how to extract the exact order since it "composes" the options of the same string_key into a map with value_type == vector< string > (in my case).
By iterating over the map, the order is lost. Is there a way to not duplicate the parsing, but have a function called (perhaps a callback) every time an option is parsed? I couldn't find documentation in this direction. Any other suggestions?
To convince you that I'm not making this up, here's what I have so far:
namespace bpo = boost::program_options;
std::vector<std::string> args_cat, args_grid, args_web;
bpo::options_description desc("Program options:");
desc.add_options()
.operator ()("help,h","Displays this help message.")
.operator ()("caterpillar,c",bpo::value< std::vector<std::string> >(&args_cat)->default_value( std::vector<std::string>(1,"4,7,2"), "4,7,2" ),"Caterpillar tree with 3 parameters")
.operator ()("grid,g",bpo::value< std::vector<std::string> >(&args_grid)->default_value( std::vector<std::string>(1,"3,4"), "3,4" ),"Rectangular grid with 2 parameters")
.operator ()("web,w",bpo::value< std::vector<std::string> >(&args_web)->default_value( std::vector<std::string>(1,"3,4"), "3,4" ),"Web with 2 parameters")
;
bpo::variables_map ops;
bpo::store(bpo::parse_command_line(argc,argv,desc),ops);
bpo::notify(ops);
if((argc < 2) || (ops.count("help"))) {
std::cout << desc << std::endl;
return;
}
//TODO: remove the following scope block after testing
{
typedef bpo::variables_map::iterator OptionsIterator;
OptionsIterator it = ops.options.begin(), it_end = ops.options.end();
while(it != it_end) {
std::cout << it->first << ": ";
BOOST_FOREACH(std::string value, it->second) {
std::cout << value << " ";
}
std::cout << std::endl;
++it;
}
return;
}
I realize that I could also include the type as a parameter and solve this problem trivially, e.g.:
./bin/make_graph --component c,4,5,1 --component w,3,3 --component c,3,1,2
but that's moving in the direction of writing a parser/validator myself (maybe even without using Boost Program Options):
./bin/make_graph --custom c,4,5,1,w,3,3,c,3,1,2
./bin/make_graph c,4,5,1,w,3,3,c,3,1,2
How would you guys recommend I do this in an elegant way?
Thanks in advance!
PS: I've searched on SO for "[boost] +sequence program options" and "[boost-program-options] +order" (and their variants) before posting this, so I apologize in advance if this turns out to be a duplicate.
Since posting the question, I did some digging and have a "hack" that works with the existing examples I had above.
bpo::parsed_options p_ops = bpo::parse_command_line(argc,argv,desc);
typedef std::vector< bpo::basic_option<char> >::iterator OptionsIterator;
OptionsIterator it = p_ops.options.begin(), it_end = p_ops.options.end();
while(it != it_end) {
std::cout << it->string_key << ": ";
BOOST_FOREACH(std::string value, it->value) {
std::cout << value << " ";
}
std::cout << std::endl;
++it;
}
The reason I call it a hack is because it accesses all arguments as strings, and one would have to extract the types from it much like bpo::variables_map does with the .as<T>() member function. EDIT: It also accesses a member of the options struct directly.
How about this:
./bin/make_graph c,4,5,1 c,3,1,2 w,3,3
Where "c,4,5,1", "c,3,1,2" and "w,3,3" are positional arguments which are stored (in order) in a std::vector<std::string> (just like --input-file in this tutorial) . Then use Boost.Tokenizer or boost::algorithm::split to extract the subtokens from each argument string.
If the graphs can be complex, you should consider making it possible for the user to specify an input file that contains the graph parameters. Boost.Program_Options can parse a user config file that uses the same syntax as the command line options.