Problem: writing to a file with C and reading with C++

[Keith Dewell]

Greetings!

My current job has brought me back to working in C++ which I haven't
used since school days. The solution to my problem may be trivial but
I have struggled with it for the last two days and would appreciate
this group's helpful expertise.

My problem may be related to mixing the C and C++ languages together.
Namely, I have a struct which I cannot change (as legacy code)
similiar to this (I will change the names throughout as I am working
on a government project):

typedef struct {
char firstMember[16];
char secondMember[16];
char thirdMember[96];
unsigned long count;
} my_Struct_T;

Instances of this struct are written to a file using fprintf like
this:

if (file_p) fprintf(file_p, MY_MACRO_FORMAT);

where MY_MACRO_FORMAT is defined like this:

#define MY_MACRO_FORMAT \
"%-12.12s\t%-8.8s\t%6d\t%-56.56s\t", \
aStruct.firstMember,aStruct.secondMember,aStruct.count,Struct.thirdMember

(aStruct being an instance of my_Struct_T)

The user selects the written file using a dialog box from the GUI in
order to read-in this data. To do this I use C++ code as follows:

std::vector<std::string> dataFromFile;
std::string aString;

//actually passed-in from another method
std::ifstream inputFileStream(&filename[0], std::ios::in);

while(std::getline(inputFileStream[0], aString, '\t'))
dataFromFile.push_back(aString);

Now the problem ...

Stepping through the code in the debugger as values are written to the
vector, I see that the values for aStruct.thirdMember are always
garbage unless it contains 15 or less characters.

BTW, the values always look good in the file to which I have written,
regardless of the number of characters for thirdMember. (I confirm
this with a text editor.)

Please offer a solution where I can still take a C++ approach (instead
of C) as that is my preference, while not changing the legacy code
(namely, the struct above).

Much Thanks,
Keith

 

[Kevin Goodsell]

Greetings!

My current job has brought me back to working in C++ which I haven't
used since school days. The solution to my problem may be trivial but
I have struggled with it for the last two days and would appreciate
this group's helpful expertise.

My problem may be related to mixing the C and C++ languages together.
Namely, I have a struct which I cannot change (as legacy code)
similiar to this (I will change the names throughout as I am working
on a government project):

typedef struct {
char firstMember[16];
char secondMember[16];
char thirdMember[96];
unsigned long count;
} my_Struct_T;

Instances of this struct are written to a file using fprintf like
this:

if (file_p) fprintf(file_p, MY_MACRO_FORMAT);

where MY_MACRO_FORMAT is defined like this:

#define MY_MACRO_FORMAT \
"%-12.12s\t%-8.8s\t%6d\t%-56.56s\t", \
aStruct.firstMember,aStruct.secondMember,aStruct.count,Struct.thirdMember

(aStruct being an instance of my_Struct_T)

The user selects the written file using a dialog box from the GUI in
order to read-in this data. To do this I use C++ code as follows:

std::vector<std::string> dataFromFile;
std::string aString;

//actually passed-in from another method
std::ifstream inputFileStream(&filename[0], std::ios::in);

What is 'filename'? If it's a std::string, this is very bad (use
filename.c_str() instead). If it's a char* or char[] this is
unnecessary. The only case I can think of where you'd want to do this is

while(std::getline(inputFileStream[0], aString, '\t'))
dataFromFile.push_back(aString);

It's hard to say what this will do without knowing something about the
format of the file. Those char[] members you wrote to the file, might
they have tab characters in them?

Now the problem ...

Stepping through the code in the debugger as values are written to the
vector, I see that the values for aStruct.thirdMember are always
garbage unless it contains 15 or less characters.

BTW, the values always look good in the file to which I have written,
regardless of the number of characters for thirdMember. (I confirm
this with a text editor.)

Please offer a solution where I can still take a C++ approach (instead
of C) as that is my preference, while not changing the legacy code
(namely, the struct above).

Your problem does not appear to have anything to do with "mixing C and
C++." There is nothing obviously wrong here (other than the
'&filename[0]' thing). Try posting a complete, minimal program that
demonstrates the problem, as well as some input to test it with.

-Kevin

 

[=?iso-8859-1?Q?Juli=E1n?= Albo]

Keith Dewell escribió:

Stepping through the code in the debugger as values are written to the
vector, I see that the values for aStruct.thirdMember are always
garbage unless it contains 15 or less characters.
BTW, the values always look good in the file to which I have written,
regardless of the number of characters for thirdMember. (I confirm
this with a text editor.)

If I understand you correctly the program works fine, the problem is how
to see correctly a non zero terminated array of char in the debugger. To
get help about that, post in a group about your development tools.

Regards.

 

[Keith Dewell]

Keith Dewell escribi? :
sí


If I understand you correctly the program works fine, the problem is how
to see correctly a non zero terminated array of char in the debugger. To
get help about that, post in a group about your development tools.

Regards.

Thanks for responding. I am using MS Visual Studio 7.0 (a.k.a. MS
Visual C++ .NET) and watch the vector become populated by using:

dataFromFile._Myfirst, 10

in the watch window of the debugger.

The other members of the struct show good data in the debugger, only
thirdMember shows garbage. This is why I believe this is an I/O
problem since the data looks good in the file I am reading and this is
the only variable where I see garbage; all the other members of the
struct show good data in the debugger as they come into the vector.

Regards,
Keith

 

[=?iso-8859-1?Q?Juli=E1n?= Albo]

Keith Dewell escribió:

The other members of the struct show good data in the debugger, only
thirdMember shows garbage. This is why I believe this is an I/O
problem since the data looks good in the file I am reading and this is
the only variable where I see garbage; all the other members of the
struct show good data in the debugger as they come into the vector.

Forget the debugger for one moment, and check the chars indvidually in
the program, then you will know if it is a problem with the debugger or
not.

If it is a program with the debugger ask in a windows programming group,
debuggers and compiler specific things are off-topic here.

Regards.

 

[Keith Dewell]

Greetings!

My current job has brought me back to working in C++ which I haven't
used since school days. The solution to my problem may be trivial but
I have struggled with it for the last two days and would appreciate
this group's helpful expertise.

My problem may be related to mixing the C and C++ languages together.
Namely, I have a struct which I cannot change (as legacy code)
similiar to this (I will change the names throughout as I am working
on a government project):

typedef struct {
char firstMember[16];
char secondMember[16];
char thirdMember[96];
unsigned long count;
} my_Struct_T;

Instances of this struct are written to a file using fprintf like
this:

if (file_p) fprintf(file_p, MY_MACRO_FORMAT);

where MY_MACRO_FORMAT is defined like this:

#define MY_MACRO_FORMAT \
"%-12.12s\t%-8.8s\t%6d\t%-56.56s\t", \
aStruct.firstMember,aStruct.secondMember,aStruct.count,Struct.thirdMember

(aStruct being an instance of my_Struct_T)

The user selects the written file using a dialog box from the GUI in
order to read-in this data. To do this I use C++ code as follows:

std::vector<std::string> dataFromFile;
std::string aString;

//actually passed-in from another method
std::ifstream inputFileStream(&filename[0], std::ios::in);

What is 'filename'? If it's a std::string, this is very bad (use
filename.c_str() instead). If it's a char* or char[] this is
unnecessary. The only case I can think of where you'd want to do this is
if filename just happened to be a null-terminated vector<char>, which is
unusual.

'filename' is an instance of FXString from the FOX Toolkit
(http://www.fox-toolkit.org/); in our shop we use FOX instead of MFC
or whatever for GUI development. We use a dialog box for the user to
give the filename and this code uses FOX. I agree that '&filename[0]'
is ugly, but it is the only way I have found, so far, that works.
FXString does not understand c_str(). Anyway, this doesn't pertain to
my problem since this code successfully attaches the stream to the
file. But thanks for your input!

while(std::getline(inputFileStream[0], aString, '\t'))
dataFromFile.push_back(aString);

It's hard to say what this will do without knowing something about the
format of the file. Those char[] members you wrote to the file, might
they have tab characters in them?

The values which the char[] members hold onto are just words like,
"The quick brown fox jumped over the lazy dog" -- or whatever words
may be found in a government project. The only whitespace
permissable is a blank. Between the struct members I insert a tab
delimiter for later parsing (see MY_MACRO_FORMAT above).

Your problem does not appear to have anything to do with "mixing C and
C++."

The reason I thought it may have to do with the mixing of languages is
that I write to the file with C and read the file with C++ where:

1)The only problem-child is the thirdMember, the first and second
members never take in garbage values and the only difference is the
size of the arrays in their definition, i.e., char firstMember[16] and
char secondMember[16] versus char thirdMember[96]. AND ...

2)The problem seems to have something to do with the "magical"
number 16, because if the thirdMember has < 16 characters then it
takes no garbage from the stream and the values look good.

There is nothing obviously wrong here (other than the
'&filename[0]' thing). Try posting a complete, minimal program that
demonstrates the problem, as well as some input to test it with.

-Kevin

I started rewriting a minimal program but found it more involved than
expected; partly because of the presence of FOX, partly because I need
to change names and data for security reasons, and partly because of
the overarching functionality which crosses over a few different
classes. I haven't given-up on this yet and it may be a good approach
for troubleshooting anyway. But I thought I should go ahead and
respond to your post with the hope of presenting my problem more
clearly.

Thanks for all of your help,
Keith

 

[Kevin Goodsell]

I started rewriting a minimal program but found it more involved than
expected;

Well, I did write a minimal program based on what you posted:

#include <stdio.h>
#include <iostream>
#include <fstream>
#include <vector>
#include <string>

using namespace std;

typedef struct {
char firstMember[16];
char secondMember[16];
char thirdMember[96];
unsigned long count;
} my_Struct_T;


#define MY_MACRO_FORMAT \
"%-12.12s\t%-8.8s\t%6d\t%-56.56s\t", \
aStruct.firstMember,aStruct.secondMember,aStruct.count,aStruct.thirdMember


int main()
{
FILE *fp = fopen("test.txt", "w");
if (fp == 0)
{
cerr << "could not open test.txt" << endl;
return 0;
}

my_Struct_T aStruct = {"first", "second",
"third is very very very very loooooonnnnnng", 24};
if (fp) fprintf(fp, MY_MACRO_FORMAT);

fclose(fp);

ifstream inputFileStream("test.txt");

std::vector<std::string> dataFromFile;
std::string aString;

while(std::getline(inputFileStream, aString, '\t'))
dataFromFile.push_back(aString);

return 0;
}


This works exactly as expected. I also noticed a few more errors in your
code while making this:

#define MY_MACRO_FORMAT \
"%-12.12s\t%-8.8s\t%6d\t%-56.56s\t", \
aStruct.firstMember,aStruct.secondMember,aStruct.count,Struct.thirdMember
^^^^^^

'Struct.thirdMember' should presumably be 'aStruct.thirdMember'.

And here:

std::ifstream inputFileStream(&filename[0], std::ios::in);

while(std::getline(inputFileStream[0], aString, '\t'))
dataFromFile.push_back(aString);

'inputFileStream[0]' should presumably NOT have the '[0]'.

-Kevin

 

[Keith Dewell]

Keith Dewell escribi :


Forget the debugger for one moment, and check the chars indvidually in
the program, then you will know if it is a problem with the debugger or
not.

If it is a program with the debugger ask in a windows programming group,
debuggers and compiler specific things are off-topic here.

Regards.

Julián,

You were right(!), the problem was related to the debugger and also to
std::string. Here is a quote from Thore B. Karlsen on another thread
dated 2003-01-03 07:24:05 PST:

"The string class in VC++ 7.0 has a short string optimization for
strings of 16 characters or less. Anything with this length is stored
in a small buffer in the string object. For longer strings, memory is
allocated. What you're seeing is probably just the internal buffer."

Hence the "magic" number 16 I was encountering. BTW, Mr. Karlsen has
provided this solution for at least 3 different posters. And though
debuggers are off-topic I thought I should post this here just so
everyone can see the resolution to my problem.

Thanks Julián (and Kevin) for your help!

Regards,
Keith