FROZEN - Better project at Stricks

Buf

Experimental string buffer library.

Why ?

Because C strings are painful, unsafe and slow.

Say you're making a forum engine,
where a page is a fixed-size buffer that you fill with posts.

How would you safely build a page without truncating posts ?

The hard way

char page[PAGE_SZ];

// Keep track !
size_t page_len = 0;

while(1) {

    char* user = db_col("user");
    char* text = db_col("text");

    // Big enough ?
    #define POST_SZ 100

    // Better malloc and realloc ?
    char post[POST_SZ];  

    // May be truncated !
    snprintf (post, POST_SZ, "<div>%s<p>%s</p></div>", user, text);

    // Typing "strlen()" 100 times a day..
    const size_t post_len = strlen(post);

    // Check if it fits !
    // What about terminating null ? Will it fit ?
    if (page_len + post_len < PAGE_SZ) {    
        // Not sure..
        strncat (page, post, PAGE_SZ-page_len); 
        // ..about all this.
        page_len += post_len;
    } else {
        break;
    } 
}

The Buf way

Buf page = buf_new(PAGE_SZ);

while(1) {

    char* user = db_col("user");
    char* text = db_col("text");

    if (!buf_append (page, "<div>%s<p>%s</p></div>", user, text)) {
        break;
    }
}

Principle

The Buf type is an opaque pointer to a struct.
This struct ends with a flexible array :

struct Buf_s *Buf;

struct Buf_s {
    uint32_t cap;  // capacity
    uint32_t len;  // current length
    unsigned char data[];  // null-terminated string
}

Making *Buf a contiguous chunk of memory.
And you still get your good old C-string through the buf_data() accessor.

Quick sample

The example folder implements the 'forum'.
It reads rows from a mock db, and renders them to a page of limited size.
When the next post won't fit, we flush the page and start anew.

make && cd example && ./forum

API

buf_new

Buf buf_new (const size_t cap);

Allocate a fresh Buf of capacity cap.
To release, simply free(buf);

buf_append

int buf_append (Buf buf, const char* fmt, ...);

Append a formatted c-string to buf.
If new data would exceed capacity, buf stays unmodified.
Returns: change in length, or zero on failure.

buf_write

int buf_write (Buf buf, const char* fmt, ...);

Write a formatted c-string at start of buf.
If new data would exceed capacity, buf stays unmodified.
Returns: new length, or zero on failure.

buf_dup

Buf buf_dup (const Buf buf);

Make a clone.

buf_resize

bool buf_resize (Buf* pbuf, const size_t newcap);

Change capacity.
If lowered below length, data will be truncated.

buf_reset

void buf_reset (Buf buf);

Set data length to zero.

Accessors

Capacity
size_t buf_cap (const Buf buf);

Current length
size_t buf_len (const Buf buf);

Data string
const char* buf_data (const Buf buf);

Usage

<pre> // app.c #include &lt;stdio.h&gt; #include "buf.h" int main() { char name[] = "Bob"; Buf msg = buf_new(100); buf_append (msg, "Hello! "); buf_append (msg, "My name is %s", name); puts (buf_data(msg)); //-> Hello! My name is Bob return 0; } </pre>

gcc app.c buf.o -o app

Build & unit-test

make && make check

TODO

• utf8 ?
• Add methods like split() ?
• Make separate FlexBuf lib with auto-resize ?