test phonetics

include/hammy/refdb.h

@@ -21,6 +21,9 @@
 // Longest code in the morse table is '$' ('...-..-'), seven characters.
 #define HAMMY_MORSE_MAX 16
 
+// Longest Phonetic code. Fuck it, 32 characters
+#define HAMMY_PHONETIC_MAX 32
+
 // The current longest string in the qcodes table is 42; 128 is pretty generous. TODO: Change this if the qcodes table ever changes
 #define HAMMY_QCODE_MAX 128
 
@@ -37,6 +40,7 @@ struct hammy_refdb_t {
     sqlite3_stmt* stPrefix;
     sqlite3_stmt* stMorse;
     sqlite3_stmt* stQCode;
+    sqlite3_stmt* stPhonetic;
     sqlite3_stmt* stFreqMain;
     sqlite3_stmt* stFreqIaru;
     sqlite3_stmt* stFreqNearest;
@@ -46,6 +50,9 @@ struct hammy_refdb_t {
 
     char morseCode[HAMMY_MORSE_MAX]; // Scratch for the last hammy_refdb_get_morse() hit
 
+    char phoneticCode[HAMMY_PHONETIC_MAX];
+    char phoneticCodePronunciation[HAMMY_PHONETIC_MAX];
+
     char qcodeQuestion[HAMMY_QCODE_MAX]; // Scratch for the last hammy_refdb_get_qcode() hit
     char qcodeAnswer[HAMMY_QCODE_MAX]; // Scratch for the last hammy_refdb_get_qcode() hit
 
@@ -154,6 +161,13 @@ bool hammy_refdb_dxcc(hammy_refdb_t* db, const char* callsign, hammy_dxcc_t* out
 // the NEXT call on the same handle - copy it if it has to outlive that.
 bool hammy_refdb_get_morse(hammy_refdb_t* db, char c, const char** out);
 
+// Looks up a character in the Phonetic table. Case-insensitive; non-ASCII bytes
+// never match. Returns false if not found, leaving *out untouched.
+//
+// On a hit *out points at storage owned by the refdb and is only valid until
+// the NEXT call on the same handle - copy it if it has to outlive that.
+bool hammy_refdb_get_phonetic(hammy_refdb_t* db, char c, const char** out, const char** outPronunciation);
+
 // Looks up a QSO code in the qcodes table. Case-insensitive; non-ASCII bytes
 // never match. Returns false if not found, leaving *out untouched.
 //

src/commands/morse.c

@@ -73,7 +73,7 @@ void hammy_cmd_morse(const hammy_job_t* job, struct discord* client, hammy_refdb
     // Get the text argument from the job
     const char* text = hammy_job_get_arg(job, "text");
     if (!text) {
-        hammy_job_respond(job, client, "Error", "No text provided for Morse code conversion.", true);
+        hammy_job_respond(job, client, "Text Missing!", "No Text provided for Morse code conversion.", true);
         return;
     }
 

src/commands/phonetic.c

@@ -0,0 +1,142 @@
+#include <concord/discord.h>
+#include <inttypes.h>
+#include <stdio.h>
+#include <string.h>
+
+#include <hammy/command.h>
+#include <hammy/job.h>
+#include <hammy/refdb.h>
+
+// The reply goes out as an embed description, which Discord caps at 4096
+// characters, so the two halves get separate budgets that still leave room for
+// the labels between them. Capping at all is what keeps an unbounded,
+// user-controlled VLA off the stack: a command option runs to 6000 characters,
+// and every one of them can expand to eight.
+#define HAMMY_PHONETIC_ECHO_MAX 512
+#define HAMMY_PHONETIC_CODE_MAX 3072
+#define HAMMY_PHONETIC_TRUNCATED " ... (truncated)"
+
+// Appends one token, space-separated from whatever is already in the buffer and
+// optionally preceded by a word-gap slash. All or nothing: returns false and
+// leaves the buffer untouched when the whole thing would not fit, so the caller
+// can stop without a half-written character on the end.
+//
+// len counts bytes excluding the terminator, and the buffer stays terminated
+// throughout - which is what the strcat-onto-uninitialised-stack version this
+// replaces got wrong: strcat needs a terminated destination to start from, and
+// an uninitialised one made the reply random bytes that Discord rejected as
+// invalid JSON.
+static bool phonetic_append(char* buf, size_t cap, size_t* len, bool gap, const char* token) {
+    size_t n = strlen(token);
+    size_t need = (*len > 0 ? 1 : 0) + (gap ? 2 : 0) + n;
+
+    if (*len + need + 1 > cap) { return false; }
+
+    if (*len > 0) { buf[(*len)++] = ' '; }
+
+    if (gap) {
+        buf[(*len)++] = '/';
+        buf[(*len)++] = ' ';
+    }
+
+    memcpy(buf + *len, token, n);
+    *len += n;
+    buf[*len] = '\0';
+
+    return true;
+}
+
+static bool phonetic_is_gap(char c) {
+    return c == ' ' || c == '\t' || c == '\n' || c == '\r';
+}
+
+// How many bytes of s can be kept without exceeding max OR splitting a UTF-8
+// sequence. Half a sequence would make the JSON body invalid and cost the whole
+// reply, so the cut walks back to the nearest lead byte.
+//
+// Only the echoed input needs this - the generated code is all ASCII.
+static size_t phonetic_utf8_trim(const char* s, size_t max) {
+    size_t n = strlen(s);
+    if (n <= max) { return n; }
+
+    // Continuation bytes are 10xxxxxx. Dropping back past them lands either on
+    // ASCII or on the lead byte of the sequence being cut, and that lead byte
+    // is then excluded too, since the kept range is [0, n).
+    n = max;
+    while (n > 0 && ((unsigned char)s[n] & 0xC0) == 0x80) { n--; }
+
+    return n;
+}
+
+// Instant command: runs on the gateway thread, sends a fresh response.
+void hammy_cmd_phonetic(const hammy_job_t* job, struct discord* client, hammy_refdb_t* refdb) {
+    // Get the text argument from the job
+    const char* text = hammy_job_get_arg(job, "callsign");
+    if (!text) {
+        hammy_job_respond(job, client, "Callsign Missing!", "No Callsign provided for Phonetic conversion.", true);
+        return;
+    }
+
+    // Echoed back trimmed rather than whole: sizing a buffer from the option
+    // would put an unbounded, user-controlled allocation on the stack.
+    char echo[HAMMY_PHONETIC_ECHO_MAX + sizeof(HAMMY_PHONETIC_TRUNCATED)];
+    size_t echoLen = morse_utf8_trim(text, HAMMY_PHONETIC_ECHO_MAX);
+
+    memcpy(echo, text, echoLen);
+    echo[echoLen] = '\0';
+
+    if (text[echoLen] != '\0') {
+        memcpy(echo + echoLen, HAMMY_PHONETIC_TRUNCATED, sizeof(HAMMY_PHONETIC_TRUNCATED));
+    }
+
+    char phonetic[HAMMY_PHONETIC_CODE_MAX + sizeof(HAMMY_PHONETIC_TRUNCATED)];
+    phonetic[0] = '\0';
+
+    // Room held back so the truncation note always fits.
+    const size_t cap = sizeof(phonetic) - (sizeof(HAMMY_PHONETIC_TRUNCATED) - 1);
+
+    size_t len = 0;
+    bool pendingGap = false;
+    bool truncated = false;
+
+    for (const char* p = text; *p; p++) {
+        // The morse table has no row for whitespace, and rendering it as an
+        // unknown character would lose the word boundaries. Held rather than
+        // emitted on the spot so leading, trailing and repeated spaces do not
+        // stack up slashes.
+        if (morse_is_gap(*p)) {
+            pendingGap = (len > 0);
+            continue;
+        }
+
+        const char* code = NULL;
+        const char* codePronunciation = NULL;
+        if (!hammy_refdb_get_phonetic(refdb, *p, &code, &codePronunciation)) {
+            code = "?"; // Handle unknown characters
+        }
+
+        if (!phonetic_append(phonetic, cap, &len, pendingGap, code)) {
+            truncated = true;
+            break;
+        }
+
+        pendingGap = false;
+    }
+
+    // All whitespace, or an empty string: Discord refuses an empty body, so
+    // there is nothing to send back.
+    if (len == 0) {
+        hammy_job_respond(job, client, "Error", "Nothing to convert.", true);
+        return;
+    }
+
+    if (truncated) {
+        memcpy(phonetic + len, HAMMY_PHONETIC_TRUNCATED, sizeof(HAMMY_PHONETIC_TRUNCATED));
+    }
+
+    // Reply with the converted Phonetic code, alongside the text it came from.
+    char body[sizeof(echo) + sizeof(phonetic) + 32];
+    snprintf(body, sizeof(body), "Original text: `%s`\nPhonetic: `%s`", echo, phonetic);
+
+    hammy_job_respond(job, client, "Phonetic Code Conversion", body, false);
+}

src/hammy/command.c

@@ -10,6 +10,7 @@ void hammy_cmd_ping(const hammy_job_t* job, struct discord* client, hammy_refdb_
 void hammy_cmd_morse(const hammy_job_t* job, struct discord* client, hammy_refdb_t* refdb);
 void hammy_cmd_freq(const hammy_job_t* job, struct discord* client, hammy_refdb_t* refdb);
 void hammy_cmd_q(const hammy_job_t* job, struct discord* client, hammy_refdb_t* refdb);
+void hammy_cmd_phonetic(const hammy_job_t* job, struct discord* client, hammy_refdb_t* refdb);
 
 // TODO: Temporary - move to a proper options system; e.g. a vector of heap-allocated options or something. For now, just point at static storage.
 static struct discord_application_command_option morse_opts[] = {
@@ -44,6 +45,16 @@ static struct discord_application_command_options q_opts_struct = {
     .array = q_opts
 };
 
+static struct discord_application_command_option phonetic_opts[] = {
+    { .type = DISCORD_APPLICATION_OPTION_STRING, .name = "callsign",
+      .description = "Callsign to convert to Phonetics", .required = true },
+};
+
+static struct discord_application_command_options phonetic_opts_struct = {
+    .size = 1,
+    .array = phonetic_opts
+};
+
 // The command table. Pure data - no allocation, no lifetime, no destructor.
 // Copying an entry into the bot's vector is a plain struct copy, and the vector
 // may be created with a NULL destructor hook.
@@ -53,7 +64,8 @@ static const hammy_command_t hammy_builtin_commands[] = {
     { "ping", "Check whether Hammy is alive and how fast it is responding", NULL, &hammy_cmd_ping, true },
     { "morse", "Text to and from Morse code", &morse_opts_struct, &hammy_cmd_morse, true },
     { "freq", "Band, segment and who can transmit", &freq_opts_struct, &hammy_cmd_freq, true },
-    { "q", "Convert Q-Code to the corresponding question and answer", &q_opts_struct, &hammy_cmd_q, true }
+    { "q", "Convert Q-Code to the corresponding question and answer", &q_opts_struct, &hammy_cmd_q, true },
+    { "phonetic", "Convert Callsign to Phonetics", &phonetic_opts_struct, &hammy_cmd_phonetic, true }
 
     // Tier 0 commands go here as they land. All pure computation, so instant:
     //   { "grid",  "Maidenhead locator conversions", &grid_opts,  &hammy_cmd_grid,  true },

src/hammy/refdb.c

@@ -90,6 +90,10 @@ static const char SQL_COUNTRY_KNOWN[] =
 static const char SQL_QCODE[] =
     "SELECT question, answer FROM qcodes WHERE code = UPPER(?1) LIMIT 1";
 
+// Resolve character to phonetic
+static const char SQL_PHONETIC[] =
+    "SELECT word, pronunciation FROM phonetics WHERE letter = UPPER(?1) LIMIT 1"
+
 // Authorizer: the connection may read, and nothing else.
 //
 // SQLITE_OPEN_READONLY protects the MAIN database file. It does not stop
@@ -143,6 +147,7 @@ static const hammy_stmt_def_t HAMMY_STATEMENTS[] = {
     HAMMY_STMT(stPrefix,       SQL_PREFIX),
     HAMMY_STMT(stMorse,        SQL_MORSE),
     HAMMY_STMT(stQCode,        SQL_QCODE),
+    HAMMY_STMT(stPhonetic,     SQL_PHONETIC),
     HAMMY_STMT(stFreqMain,     SQL_FREQ_MAIN),
     HAMMY_STMT(stFreqIaru,     SQL_FREQ_IARU),
     HAMMY_STMT(stFreqNearest,  SQL_FREQ_NEAREST),
@@ -451,6 +456,48 @@ bool hammy_refdb_get_morse(hammy_refdb_t* db, char c, const char** out) {
     return hit;
 }
 
+bool hammy_refdb_get_phonetic(hammy_refdb_t* db, char c, const char** out, const char** outPronunciation) {
+    if (!db || !out) { return false; }
+
+    sqlite3_stmt* const needed[] = { db->stPhonetic };
+    if (!stmts_ready("phonetic", needed, 1)) { return false; }
+
+    // The table is ASCII (ITU-R M.1677-1 has no non-Latin extensions), and a
+    // single byte out of a multi-byte UTF-8 sequence is not valid text to bind,
+    // so anything with the high bit set is a miss without touching SQLite.
+    if ((unsigned char)c & 0x80u) { return false; }
+
+    // Run the query with the character as a string. SQLite's UPPER() handles case-insensitivity.
+    // SQLITE_STATIC is safe because the statement is reset before this returns,
+    // so key never outlives the binding that points at it.
+    char key[2] = { c, '\0' };
+
+    sqlite3_reset(db->stPhonetic);
+    sqlite3_clear_bindings(db->stPhonetic);
+    sqlite3_bind_text(db->stPhonetic, 1, key, 1, SQLITE_STATIC);
+
+    bool hit = false;
+
+    if (sqlite3_step(db->stPhonetic) == SQLITE_ROW) {
+        const unsigned char* code = sqlite3_column_text(db->stPhonetic, 0);
+        const unsigned char* codePronunciation = sqlite3_column_text(db->stPhonetic, 1);
+
+        if (code && codePronunciation) {
+            // Copied rather than handed back directly: the column pointer dies
+            // at the reset below, and the caller has no way to know that.
+            snprintf(db->phoneticCode, sizeof(db->phoneticCode), "%s", (const char*)code);
+            *out = db->phoneticCode;
+            snprintf(db->phoneticCodePronunciation, sizeof(db->phoneticCodePronunciation), "%s", (const char*)codePronunciation);
+            *outPronunciation = db->phoneticCodePronunciation;
+            hit = true;
+        }
+    }
+
+    sqlite3_reset(db->stPhonetic);
+
+    return hit;
+}
+
 bool hammy_refdb_get_qcode(hammy_refdb_t* db, const char* code, const char** outQuestion, const char** outAnswer) {
     if (!db || !code || !outQuestion || !outAnswer) { return false; }