Add the EN_loadpatternfile function to load pattern data from a .pat file into INP

[Mariosmsk]

No description provided.

[LRossman]

I took the liberty of re-writing your EN_loadpatternfile function. It avoids reallocating memory each time a new line is read from the file and uses EPANET's own getfloat function in place of the unreliable atof function to convert strings to numbers. Please see the attached file.

EN_loadpatternfile.txt

[Mariosmsk]

It's good that you took the initiative, @LRossman. Feel free to make any changes as needed. Thank you! Also, we now need to add the function EN_savepatternfile.

[LRossman]

I was thinking, why have the function work only for a new pattern and not for an existing one as well. If you agree with this, then the code modification needed is shown below:

    file = fopen(filename, "r");
    if (file == NULL) return 302;

    // Add a new pattern if it doesn't already exist
    err = EN_getpatternindex(p, id, &i);
    if (err == 205) {
        if ((err = EN_addpattern(p, id)) != 0) {
            fclose(file);
            return err;
        }
        i = p->network.Npats;
    }

    // Read pattern values

[Mariosmsk]

Yes, it's a good idea to use an existing pattern ID not only for new patterns, and I agree.

Regarding the getfloat function, there seems to be an issue in EN_loadpatternfile (error 205). So, I made the changes below in getfloat; however, I'm not sure if I have broken something else. What do you think?

from

{
    char *endptr;
    *y = (double)strtod(s, &endptr);
    if (*endptr > 0) return 0;
    return 1;
}

to

{
    char *endptr;
    *y = strtod(s, &endptr);

    if (endptr == s || (*endptr != '\0' && !isspace((unsigned char)*endptr))) {
        return 0;
    }
    return 1;
}

[LRossman]

I would rather not change getfloat since it's used extensively when processing an input file. If EN_loadpatternfile is returning an error 205 then it seems that either:

1. The 205 error from EN_getpatternindex is not being reset to 0 before a new pattern is created.
2. The values array passed into EN_setpattern is NULL implying that there's a problem with reallocating its size as the file is read.

Neither of these involves getfloat. I will download your code onto my machine and try to get a better understanding of why its failing.

[Mariosmsk]

I checked it, and the problem is that the len=0. I believe the code skipped all the lines, so when I changed the line with getfloat, it worked.

[LRossman]

OK, good. My preference is to change the following line:

    err = EN_setpattern(p, i, values, len);

to

    if (len > 0) err = EN_setpattern(p, i, values, len);

[Mariosmsk]

But if we add len > 0, I think it doesn't do anything to set the patterns. The problem is that the code doesn't execute the line len++; so the length stays at zero. Maybe I am doing something wrong. I think it's good to test it.

[LRossman]

The way I read the code is that if a line is blank or doesn't contain a valid number then getfloat will return 0 and that line will be skipped. If that happens for all lines (or the file is empty) then the values array stays at NULL and len remains at 0. If you try to call EN_setpattern with a NULL ``values array it will return with error 205. Therefore testing if len > 0 before calling EN_setpattern will avoid that happening. And if len = 0 then the pattern's multipliers will remain unchanged.

But you're right in saying that the code needs to be tested.

[LRossman]

I think I found the problem. When getfloat is applied to line it sees the newline character at the end and therefore thinks the number contained in the line is invalid. (The reason we shouldn't simply use atof is that it won't catch an invalid number caused by a typo, e.g. it would read "123X.45" as "123".) When getfloat is used to read values from the input file lines have already been parsed into string tokens without any newline characters so it works fine in that case.

The fix is as follows:

    char *tok;
    ....
    // Read pattern values
    while (fgets(line, sizeof(line), file) != NULL) {
    
        // Skip lines that don't contain valid numbers
        tok = strtok(line, SEPSTR);
        if (tok == NULL) continue;
        if (!getfloat(tok, &value)) continue;

        // Resize multiplier array if it's full

I tested this updated code on a number of pattern files, some with all numbers, some with blank or comment lines and some empty, and everything seemed to work ok.

[Mariosmsk]

Yes, it looks good now. Thank you, @LRossman!