Compress all translated strings with Huffman coding.

This saves code space in builds which use link-time optimization.
The optimization drops the untranslated strings and replaces them
with a compressed_string_t struct. It can then be decompressed to
a c string.

Builds without LTO work as well but include both untranslated
strings and compressed strings.

This work could be expanded to include QSTRs and loaded strings if
a compress method is added to C. Its tracked in #531.

.gitmodules

@@ -80,3 +80,6 @@
 	path = lib/tinyusb
 	url = https://github.com/hathach/tinyusb.git
 	branch = develop
+[submodule "tools/huffman"]
+	path = tools/huffman
+	url = https://github.com/tannewt/huffman.git

extmod/machine_mem.c

@@ -42,7 +42,7 @@
 STATIC uintptr_t machine_mem_get_addr(mp_obj_t addr_o, uint align) {
     uintptr_t addr = mp_obj_int_get_truncated(addr_o);
     if ((addr & (align - 1)) != 0) {
-        nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_ValueError, "address %08x is not aligned to %d bytes", addr, align));
+        nlr_raise(mp_obj_new_exception_msg_varg(&mp_type_ValueError, translate("address %08x is not aligned to %d bytes"), addr, align));
     }
     return addr;
 }

extmod/modframebuf.c

@@ -296,7 +296,7 @@ STATIC mp_obj_t framebuf_make_new(const mp_obj_type_t *type, size_t n_args, size
         case FRAMEBUF_GS8:
             break;
         default:
-            mp_raise_ValueError("invalid format");
+            mp_raise_ValueError(translate("invalid format"));
     }
 
     return MP_OBJ_FROM_PTR(o);

extmod/modubinascii.c

@@ -35,7 +35,7 @@
 static void check_not_unicode(const mp_obj_t arg) {
 #if MICROPY_CPYTHON_COMPAT
     if (MP_OBJ_IS_STR(arg)) {
-        mp_raise_TypeError("a bytes-like object is required");
+        mp_raise_TypeError(translate("a bytes-like object is required"));
     }
 #endif
 }
@@ -87,7 +87,7 @@ mp_obj_t mod_binascii_unhexlify(mp_obj_t data) {
     mp_get_buffer_raise(data, &bufinfo, MP_BUFFER_READ);
 
     if ((bufinfo.len & 1) != 0) {
-        mp_raise_ValueError("odd-length string");
+        mp_raise_ValueError(translate("odd-length string"));
     }
     vstr_t vstr;
     vstr_init_len(&vstr, bufinfo.len / 2);
@@ -98,7 +98,7 @@ mp_obj_t mod_binascii_unhexlify(mp_obj_t data) {
         if (unichar_isxdigit(hex_ch)) {
             hex_byte += unichar_xdigit_value(hex_ch);
         } else {
-            mp_raise_ValueError("non-hex digit found");
+            mp_raise_ValueError(translate("non-hex digit found"));
         }
         if (i & 1) {
             hex_byte <<= 4;
@@ -166,7 +166,7 @@ mp_obj_t mod_binascii_a2b_base64(mp_obj_t data) {
     }
 
     if (nbits) {
-        mp_raise_ValueError("incorrect padding");
+        mp_raise_ValueError(translate("incorrect padding"));
     }
 
     return mp_obj_new_str_from_vstr(&mp_type_bytes, &vstr);

extmod/modure.c

@@ -158,7 +158,7 @@ STATIC mp_obj_t re_split(size_t n_args, const mp_obj_t *args) {
         mp_obj_t s = mp_obj_new_str_of_type(str_type, (const byte*)subj.begin, caps[0] - subj.begin);
         mp_obj_list_append(retval, s);
         if (self->re.sub > 0) {
-            mp_raise_NotImplementedError("Splitting with sub-captures");
+            mp_raise_NotImplementedError(translate("Splitting with sub-captures"));
         }
         subj.begin = caps[1];
         if (maxsplit > 0 && --maxsplit == 0) {
@@ -204,7 +204,7 @@ STATIC mp_obj_t mod_re_compile(size_t n_args, const mp_obj_t *args) {
     int error = re1_5_compilecode(&o->re, re_str);
     if (error != 0) {
 error:
-        mp_raise_ValueError("Error in regex");
+        mp_raise_ValueError(translate("Error in regex"));
     }
     if (flags & FLAG_DEBUG) {
         re1_5_dumpcode(&o->re);

main.c

@@ -128,13 +128,22 @@ const char* first_existing_file_in_list(const char ** filenames) {
     return NULL;
 }
 
+void write_compressed(const compressed_string_t* compressed) {
+    char decompressed[compressed->length];
+    decompress(compressed, decompressed);
+    serial_write(decompressed);
+}
+
 bool maybe_run_list(const char ** filenames, pyexec_result_t* exec_result) {
     const char* filename = first_existing_file_in_list(filenames);
     if (filename == NULL) {
         return false;
     }
     mp_hal_stdout_tx_str(filename);
-    mp_hal_stdout_tx_str(translate(" output:\n"));
+    const compressed_string_t* compressed = translate(" output:\n");
+    char decompressed[compressed->length];
+    decompress(compressed, decompressed);
+    mp_hal_stdout_tx_str(decompressed);
     pyexec_file(filename, exec_result);
     return true;
 }
@@ -145,11 +154,11 @@ bool run_code_py(safe_mode_t safe_mode) {
     if (serial_connected_at_start) {
         serial_write("\n");
         if (autoreload_is_enabled()) {
-            serial_write(translate("Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.\n"));
+            write_compressed(translate("Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.\n"));
         } else if (safe_mode != NO_SAFE_MODE) {
-            serial_write(translate("Running in safe mode! Auto-reload is off.\n"));
+            write_compressed(translate("Running in safe mode! Auto-reload is off.\n"));
         } else if (!autoreload_is_enabled()) {
-            serial_write(translate("Auto-reload is off.\n"));
+            write_compressed(translate("Auto-reload is off.\n"));
         }
     }
     #endif
@@ -163,7 +172,7 @@ bool run_code_py(safe_mode_t safe_mode) {
     bool found_main = false;
 
     if (safe_mode != NO_SAFE_MODE) {
-        serial_write(translate("Running in safe mode! Not running saved code.\n"));
+        write_compressed(translate("Running in safe mode! Not running saved code.\n"));
     } else {
         new_status_color(MAIN_RUNNING);
 
@@ -179,7 +188,7 @@ bool run_code_py(safe_mode_t safe_mode) {
         if (!found_main){
             found_main = maybe_run_list(double_extension_filenames, &result);
             if (found_main) {
-                serial_write(translate("WARNING: Your code filename has two extensions\n"));
+                write_compressed(translate("WARNING: Your code filename has two extensions\n"));
             }
         }
         stop_mp();
@@ -218,37 +227,37 @@ bool run_code_py(safe_mode_t safe_mode) {
 
             if (!serial_connected_at_start) {
                 if (autoreload_is_enabled()) {
-                    serial_write(translate("Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.\n"));
+                    write_compressed(translate("Auto-reload is on. Simply save files over USB to run them or enter REPL to disable.\n"));
                 } else {
-                    serial_write(translate("Auto-reload is off.\n"));
+                    write_compressed(translate("Auto-reload is off.\n"));
                 }
             }
             // Output a user safe mode string if its set.
             #ifdef BOARD_USER_SAFE_MODE
             if (safe_mode == USER_SAFE_MODE) {
                 serial_write("\n");
-                serial_write(translate("You requested starting safe mode by "));
+                write_compressed(translate("You requested starting safe mode by "));
                 serial_write(BOARD_USER_SAFE_MODE_ACTION);
                 serial_write("\n");
-                serial_write(translate("To exit, please reset the board without "));
+                write_compressed(translate("To exit, please reset the board without "));
                 serial_write(BOARD_USER_SAFE_MODE_ACTION);
                 serial_write("\n");
             } else
             #endif
             if (safe_mode != NO_SAFE_MODE) {
                 serial_write("\n");
-                serial_write(translate("You are running in safe mode which means something really bad happened.\n"));
+                write_compressed(translate("You are running in safe mode which means something really bad happened.\n"));
                 if (safe_mode == HARD_CRASH) {
-                    serial_write(translate("Looks like our core CircuitPython code crashed hard. Whoops!\n"));
+                    write_compressed(translate("Looks like our core CircuitPython code crashed hard. Whoops!\n"));
-                    serial_write(translate("Please file an issue here with the contents of your CIRCUITPY drive:\n"));
+                    write_compressed(translate("Please file an issue here with the contents of your CIRCUITPY drive:\n"));
                     serial_write("https://github.com/adafruit/circuitpython/issues\n");
                 } else if (safe_mode == BROWNOUT) {
-                    serial_write(translate("The microcontroller's power dipped. Please make sure your power supply provides\n"));
+                    write_compressed(translate("The microcontroller's power dipped. Please make sure your power supply provides\n"));
-                    serial_write(translate("enough power for the whole circuit and press reset (after ejecting CIRCUITPY).\n"));
+                    write_compressed(translate("enough power for the whole circuit and press reset (after ejecting CIRCUITPY).\n"));
                 }
             }
             serial_write("\n");
-            serial_write(translate("Press any key to enter the REPL. Use CTRL-D to reload."));
+            write_compressed(translate("Press any key to enter the REPL. Use CTRL-D to reload."));
         }
         if (serial_connected_before_animation && !serial_connected()) {
             serial_connected_at_start = false;
@@ -403,7 +412,7 @@ int __attribute__((used)) main(void) {
         }
         if (exit_code == PYEXEC_FORCED_EXIT) {
             if (!first_run) {
-                serial_write(translate("soft reboot\n"));
+                write_compressed(translate("soft reboot\n"));
             }
             first_run = false;
             skip_repl = run_code_py(safe_mode);

ports/atmel-samd/audio_dma.c

@@ -33,6 +33,7 @@
 #include "shared-bindings/audioio/WaveFile.h"
 
 #include "py/mpstate.h"
+#include "py/runtime.h"
 
 static audio_dma_t* audio_dma_state[AUDIO_DMA_CHANNEL_COUNT];
 
@@ -279,6 +280,10 @@ audio_dma_result audio_dma_setup_playback(audio_dma_t* dma,
         // We're likely double buffering so set up the block interrupts.
         turn_on_event_system();
         dma->event_channel = find_sync_event_channel();
+
+        if (dma->event_channel >= EVSYS_SYNCH_NUM) {
+            mp_raise_RuntimeError(translate("All sync event channels in use"));
+        }
         init_event_channel_interrupt(dma->event_channel, CORE_GCLK, EVSYS_ID_GEN_DMAC_CH_0 + dma_channel);
 
         // We keep the audio_dma_t for internal use and the sample as a root pointer because it

ports/atmel-samd/bindings/samd/Clock.c

@@ -132,9 +132,9 @@ STATIC mp_obj_t samd_clock_set_calibration(mp_obj_t self_in, mp_obj_t calibratio
     samd_clock_obj_t *self = MP_OBJ_TO_PTR(self_in);
     int ret = clock_set_calibration(self->type, self->index, mp_obj_get_int(calibration));
     if (ret == -2)
-        mp_raise_AttributeError("calibration is read only");
+        mp_raise_AttributeError(translate("calibration is read only"));
     if (ret == -1)
-        mp_raise_ValueError("calibration is out of range");
+        mp_raise_ValueError(translate("calibration is out of range"));
     return mp_const_none;
 }
 

ports/atmel-samd/common-hal/audiobusio/PDMIn.c

@@ -357,6 +357,9 @@ uint32_t common_hal_audiobusio_pdmin_record_to_buffer(audiobusio_pdmin_obj_t* se
         uint16_t* output_buffer, uint32_t output_buffer_length) {
     uint8_t dma_channel = find_free_audio_dma_channel();
     uint8_t event_channel = find_sync_event_channel();
+    if (event_channel >= EVSYS_SYNCH_NUM) {
+        mp_raise_RuntimeError(translate("All sync event channels in use"));
+    }
 
     // We allocate two buffers on the stack to use for double buffering.
     const uint8_t samples_per_buffer = SAMPLES_PER_BUFFER;

ports/atmel-samd/common-hal/audioio/AudioOut.c

@@ -211,6 +211,10 @@ void common_hal_audioio_audioout_construct(audioio_audioout_obj_t* self,
     // Find a free event channel. We start at the highest channels because we only need and async
     // path.
     uint8_t channel = find_async_event_channel();
+    if (channel >= EVSYS_CHANNELS) {
+        mp_raise_RuntimeError(translate("All event channels in use"));
+    }
+
     #ifdef SAMD51
     connect_event_user_to_channel(EVSYS_ID_USER_DAC_START_1, channel);
     #define EVSYS_ID_USER_DAC_START EVSYS_ID_USER_DAC_START_0

ports/atmel-samd/common-hal/busio/SPI.c

@@ -129,7 +129,7 @@ void common_hal_busio_spi_construct(busio_spi_obj_t *self,
     }
     #endif
     if (sercom == NULL) {
-        mp_raise_ValueError("Invalid pins");
+        mp_raise_ValueError(translate("Invalid pins"));
     }
 
     // Set up SPI clocks on SERCOM.

ports/atmel-samd/peripherals

@@ -1 +1 @@
-Subproject commit 1140ff6d7ed413aa1e2f34c5d847dcc41c924bb9
+Subproject commit 4922fb5c062c7e6e4438966b559eb4b9e0bb1ecd

py/compile.c

@@ -156,7 +156,7 @@ STATIC void compile_error_set_line(compiler_t *comp, mp_parse_node_t pn) {
     }
 }
 
-STATIC void compile_syntax_error(compiler_t *comp, mp_parse_node_t pn, const char *msg) {
+STATIC void compile_syntax_error(compiler_t *comp, mp_parse_node_t pn, const compressed_string_t *msg) {
     // only register the error if there has been no other error
     if (comp->compile_error == MP_OBJ_NULL) {
         comp->compile_error = mp_obj_new_exception_msg(&mp_type_SyntaxError, msg);
@@ -949,7 +949,7 @@ STATIC void compile_del_stmt(compiler_t *comp, mp_parse_node_struct_t *pns) {
 
 STATIC void compile_break_cont_stmt(compiler_t *comp, mp_parse_node_struct_t *pns) {
     uint16_t label;
-    const char *error_msg;
+    const compressed_string_t *error_msg;
     if (MP_PARSE_NODE_STRUCT_KIND(pns) == PN_break_stmt) {
         label = comp->break_label;
         error_msg = translate("'break' outside loop");

py/makeqstrdata.py

@@ -12,6 +12,10 @@ import sys
 import collections
 import gettext
 
+sys.path.append("../../tools/huffman")
+
+import huffman
+
 # Python 2/3 compatibility:
 #   - iterating through bytes is different
 #   - codepoint2name lives in a different module
@@ -83,9 +87,144 @@ def translate(translation_file, i18ns):
             unescaped = original
             for s in C_ESCAPES:
                 unescaped = unescaped.replace(C_ESCAPES[s], s)
-            translations.append((original, table.gettext(unescaped)))
+            translation = table.gettext(unescaped)
+            # Add in carriage returns to work in terminals
+            translation = translation.replace("\n", "\r\n")
+            translations.append((original, translation))
         return translations
 
+def compute_huffman_coding(translations, qstrs, compression_filename):
+    all_strings = [x[1] for x in translations]
+
+    # go through each qstr and print it out
+    for _, _, qstr in qstrs.values():
+        all_strings.append(qstr)
+    all_strings_concat = "".join(all_strings).encode("utf-8")
+    counts = collections.Counter(all_strings_concat)
+    # add other values
+    for i in range(256):
+        if i not in counts:
+            counts[i] = 0
+    cb = huffman.codebook(counts.items())
+    values = bytearray()
+    length_count = {}
+    renumbered = 0
+    last_l = None
+    canonical = {}
+    for ch, code in sorted(cb.items(), key=lambda x: (len(x[1]), x[0])):
+        values.append(ch)
+        l = len(code)
+        if l not in length_count:
+            length_count[l] = 0
+        length_count[l] += 1
+        if last_l:
+            renumbered <<= (l - last_l)
+        canonical[ch] = '{0:0{width}b}'.format(renumbered, width=l)
+        if chr(ch) in C_ESCAPES:
+            s = C_ESCAPES[chr(ch)]
+        else:
+            s = chr(ch)
+        print("//", ch, s, counts[ch], canonical[ch], renumbered)
+        renumbered += 1
+        last_l = l
+    lengths = bytearray()
+    for i in range(1, max(length_count) + 1):
+        lengths.append(length_count.get(i, 0))
+    print("//", values, lengths)
+    with open(compression_filename, "w") as f:
+        f.write("const uint8_t lengths[] = {{ {} }};\n".format(", ".join(map(str, lengths))))
+        f.write("const uint8_t values[256] = {{ {} }};\n".format(", ".join(map(str, values))))
+    return values, lengths
+
+def decompress(encoding_table, length, encoded):
+    values, lengths = encoding_table
+    #print(l, encoded)
+    dec = bytearray(length)
+    this_byte = 0
+    this_bit = 7
+    b = encoded[this_byte]
+    for i in range(length):
+        bits = 0
+        bit_length = 0
+        max_code = lengths[0]
+        searched_length = lengths[0]
+        while True:
+            bits <<= 1
+            if 0x80 & b:
+                bits |= 1
+
+            b <<= 1
+            bit_length += 1
+            if this_bit == 0:
+                this_bit = 7
+                this_byte += 1
+                if this_byte < len(encoded):
+                    b = encoded[this_byte]
+            else:
+                this_bit -= 1
+            if max_code > 0 and bits < max_code:
+                #print('{0:0{width}b}'.format(bits, width=bit_length))
+                break
+            max_code = (max_code << 1) + lengths[bit_length]
+            searched_length += lengths[bit_length]
+
+        v = values[searched_length + bits - max_code]
+        dec[i] = v
+    return dec
+
+def compress(encoding_table, decompressed):
+    if not isinstance(decompressed, bytes):
+        raise TypeError()
+    values, lengths = encoding_table
+    enc = bytearray(len(decompressed))
+    #print(decompressed)
+    #print(lengths)
+    current_bit = 7
+    current_byte = 0
+    for c in decompressed:
+        #print()
+        #print("char", c, values.index(c))
+        start = 0
+        end = lengths[0]
+        bits = 1
+        compressed = None
+        code = 0
+        while compressed is None:
+            s = start
+            e = end
+            #print("{0:0{width}b}".format(code, width=bits))
+            # Binary search!
+            while e > s:
+                midpoint = (s + e) // 2
+                #print(s, e, midpoint)
+                if values[midpoint] == c:
+                    compressed = code + (midpoint - start)
+                    #print("found {0:0{width}b}".format(compressed, width=bits))
+                    break
+                elif c < values[midpoint]:
+                    e = midpoint
+                else:
+                    s = midpoint + 1
+            code += end - start
+            code <<= 1
+            start = end
+            end += lengths[bits]
+            bits += 1
+            #print("next bit", bits)
+
+        for i in range(bits - 1, 0, -1):
+            if compressed & (1 << (i - 1)):
+                enc[current_byte] |= 1 << current_bit
+            if current_bit == 0:
+                current_bit = 7
+                #print("packed {0:0{width}b}".format(enc[current_byte], width=8))
+                current_byte += 1
+            else:
+                current_bit -= 1
+    if current_bit != 7:
+        current_byte += 1
+    return enc[:current_byte]
+
 def qstr_escape(qst):
     def esc_char(m):
         c = ord(m.group(0))
@@ -178,7 +317,7 @@ def make_bytes(cfg_bytes_len, cfg_bytes_hash, qstr):
     qhash_str = ('\\x%02x' * cfg_bytes_hash) % tuple(((qhash >> (8 * i)) & 0xff) for i in range(cfg_bytes_hash))
     return '(const byte*)"%s%s" "%s"' % (qhash_str, qlen_str, qdata)
 
-def print_qstr_data(qcfgs, qstrs, i18ns):
+def print_qstr_data(encoding_table, qcfgs, qstrs, i18ns):
     # get config variables
     cfg_bytes_len = int(qcfgs['BYTES_IN_LEN'])
     cfg_bytes_hash = int(qcfgs['BYTES_IN_HASH'])
@@ -191,6 +330,7 @@ def print_qstr_data(qcfgs, qstrs, i18ns):
     print('QDEF(MP_QSTR_NULL, (const byte*)"%s%s" "")' % ('\\x00' * cfg_bytes_hash, '\\x00' * cfg_bytes_len))
 
     total_qstr_size = 0
+    total_qstr_compressed_size = 0
     # go through each qstr and print it out
     for order, ident, qstr in sorted(qstrs.values(), key=lambda x: x[0]):
         qbytes = make_bytes(cfg_bytes_len, cfg_bytes_hash, qstr)
@@ -198,17 +338,23 @@ def print_qstr_data(qcfgs, qstrs, i18ns):
         total_qstr_size += len(qstr)
 
     total_text_size = 0
+    total_text_compressed_size = 0
     for original, translation in i18ns:
-        # Add in carriage returns to work in terminals
+        translation_encoded = translation.encode("utf-8")
-        translation = translation.replace("\n", "\r\n")
+        compressed = compress(encoding_table, translation_encoded)
-        for s in C_ESCAPES:
+        total_text_compressed_size += len(compressed)
-            translation = translation.replace(s, C_ESCAPES[s])
+        decompressed = decompress(encoding_table, len(translation_encoded), compressed).decode("utf-8")
-        print("TRANSLATION(\"{}\", \"{}\")".format(original, translation))
+        for c in C_ESCAPES:
-        total_text_size += len(translation)
+            decompressed.replace(c, C_ESCAPES[c])
+        #print("// \"{}\"".format(translation))
+        print("TRANSLATION(\"{}\", {}, {{ {} }}) // {}".format(original, len(translation_encoded)+1, ", ".join(["0x{:02x}".format(x) for x in compressed]), decompressed))
+        total_text_size += len(translation.encode("utf-8"))
 
     print()
     print("// {} bytes worth of qstr".format(total_qstr_size))
     print("// {} bytes worth of translations".format(total_text_size))
+    print("// {} bytes worth of translations compressed".format(total_text_compressed_size))
+    print("// {} bytes saved".format(total_text_size - total_text_compressed_size))
 
 def print_qstr_enums(qstrs):
     # print out the starter of the generated C header file
@@ -230,12 +376,15 @@ if __name__ == "__main__":
                         help='an integer for the accumulator')
     parser.add_argument('--translation', default=None, type=str,
                         help='translations for i18n() items')
+    parser.add_argument('--compression_filename', default=None, type=str,
+                        help='header for compression info')
 
     args = parser.parse_args()
 
     qcfgs, qstrs, i18ns = parse_input_headers(args.infiles)
     if args.translation:
         translations = translate(args.translation, i18ns)
-        print_qstr_data(qcfgs, qstrs, translations)
+        encoding_table = compute_huffman_coding(translations, qstrs, args.compression_filename)
+        print_qstr_data(encoding_table, qcfgs, qstrs, translations)
     else:
         print_qstr_enums(qstrs)

py/modbuiltins.c

@@ -316,7 +316,7 @@ MP_DEFINE_CONST_FUN_OBJ_KW(mp_builtin_min_obj, 1, mp_builtin_min);
 STATIC mp_obj_t mp_builtin_next(mp_obj_t o) {
     mp_obj_t ret = mp_iternext_allow_raise(o);
     if (ret == MP_OBJ_STOP_ITERATION) {
-        mp_raise_msg(&mp_type_StopIteration, "");
+        mp_raise_msg(&mp_type_StopIteration, NULL);
     } else {
         return ret;
     }

py/moduerrno.c

@@ -102,20 +102,6 @@ const mp_obj_module_t mp_module_uerrno = {
 };
 
 const char* mp_errno_to_str(mp_obj_t errno_val) {
-    // For commonly encountered errors, return human readable strings
-    if (MP_OBJ_IS_SMALL_INT(errno_val)) {
-        switch (MP_OBJ_SMALL_INT_VALUE(errno_val)) {
-            case EPERM:  return translate("Permission denied");
-            case ENOENT: return translate("No such file/directory");
-            case EIO:    return translate("Input/output error");
-            case EACCES: return translate("Permission denied");
-            case EEXIST: return translate("File exists");
-            case ENODEV: return translate("Unsupported operation");
-            case EINVAL: return translate("Invalid argument");
-            case EROFS:  return translate("Read-only filesystem");
-        }
-    }
-
     // Otherwise, return the Exxxx string for that error code
     #if MICROPY_PY_UERRNO_ERRORCODE
     // We have the errorcode dict so can do a lookup using the hash map
@@ -148,3 +134,21 @@ const char* mp_errno_to_str(mp_obj_t errno_val) {
 }
 
 #endif //MICROPY_PY_UERRNO
+
+
+// For commonly encountered errors, return human readable strings
+const compressed_string_t* mp_common_errno_to_str(mp_obj_t errno_val) {
+    if (MP_OBJ_IS_SMALL_INT(errno_val)) {
+        switch (MP_OBJ_SMALL_INT_VALUE(errno_val)) {
+            case EPERM:  return translate("Permission denied");
+            case ENOENT: return translate("No such file/directory");
+            case EIO:    return translate("Input/output error");
+            case EACCES: return translate("Permission denied");
+            case EEXIST: return translate("File exists");
+            case ENODEV: return translate("Unsupported operation");
+            case EINVAL: return translate("Invalid argument");
+            case EROFS:  return translate("Read-only filesystem");
+        }
+    }
+    return NULL;
+}

py/mperrno.h

@@ -141,5 +141,7 @@
 #endif
 
 const char* mp_errno_to_str(mp_obj_t errno_val);
+// For commonly encountered errors, return compressed human readable strings
+const compressed_string_t* mp_common_errno_to_str(mp_obj_t errno_val);
 
 #endif // MICROPY_INCLUDED_PY_MPERRNO_H

py/obj.c

@@ -86,19 +86,35 @@ void mp_obj_print_exception(const mp_print_t *print, mp_obj_t exc) {
         mp_obj_exception_get_traceback(exc, &n, &values);
         if (n > 0) {
             assert(n % 3 == 0);
-            mp_print_str(print, translate("Traceback (most recent call last):\n"));
+            // Decompress the format strings
+            const compressed_string_t* traceback = translate("Traceback (most recent call last):\n");
+            char decompressed[traceback->length];
+            decompress(traceback, decompressed);
+#if MICROPY_ENABLE_SOURCE_LINE
+            const compressed_string_t* frame = translate("  File \"%q\", line %d");
+#else
+            const compressed_string_t* frame = translate("  File \"%q\"");
+#endif
+            char decompressed_frame[frame->length];
+            decompress(frame, decompressed_frame);
+            const compressed_string_t* block_fmt = translate(", in %q\n");
+            char decompressed_block[block_fmt->length];
+            decompress(block_fmt, decompressed_block);
+
+            // Print the traceback
+            mp_print_str(print, decompressed);
             for (int i = n - 3; i >= 0; i -= 3) {
 #if MICROPY_ENABLE_SOURCE_LINE
-                mp_printf(print, translate("  File \"%q\", line %d"), values[i], (int)values[i + 1]);
+                mp_printf(print, decompressed_frame, values[i], (int)values[i + 1]);
 #else
-                mp_printf(print, translate("  File \"%q\""), values[i]);
+                mp_printf(print, decompressed_frame, values[i]);
 #endif
                 // the block name can be NULL if it's unknown
                 qstr block = values[i + 2];
                 if (block == MP_QSTR_NULL) {
                     mp_print_str(print, "\n");
                 } else {
-                    mp_printf(print, translate(", in %q\n"), block);
+                    mp_printf(print, decompressed_block, block);
                 }
             }
         }

py/obj.h

@@ -34,6 +34,8 @@
 #include "py/mpprint.h"
 #include "py/runtime0.h"
 
+#include "supervisor/shared/translate.h"
+
 // This is the definition of the opaque MicroPython object type.
 // All concrete objects have an encoding within this type and the
 // particular encoding is specified by MICROPY_OBJ_REPR.
@@ -639,9 +641,9 @@ mp_obj_t mp_obj_new_complex(mp_float_t real, mp_float_t imag);
 mp_obj_t mp_obj_new_exception(const mp_obj_type_t *exc_type);
 mp_obj_t mp_obj_new_exception_arg1(const mp_obj_type_t *exc_type, mp_obj_t arg);
 mp_obj_t mp_obj_new_exception_args(const mp_obj_type_t *exc_type, size_t n_args, const mp_obj_t *args);
-mp_obj_t mp_obj_new_exception_msg(const mp_obj_type_t *exc_type, const char *msg);
+mp_obj_t mp_obj_new_exception_msg(const mp_obj_type_t *exc_type, const compressed_string_t *msg);
-mp_obj_t mp_obj_new_exception_msg_varg(const mp_obj_type_t *exc_type, const char *fmt, ...); // counts args by number of % symbols in fmt, excluding %%; can only handle void* sizes (ie no float/double!)
+mp_obj_t mp_obj_new_exception_msg_varg(const mp_obj_type_t *exc_type, const compressed_string_t *fmt, ...); // counts args by number of % symbols in fmt, excluding %%; can only handle void* sizes (ie no float/double!)
-mp_obj_t mp_obj_new_exception_msg_vlist(const mp_obj_type_t *exc_type, const char *fmt, va_list ap); // counts args by number of % symbols in fmt, excluding %%; can only handle void* sizes (ie no float/double!)
+mp_obj_t mp_obj_new_exception_msg_vlist(const mp_obj_type_t *exc_type, const compressed_string_t *fmt, va_list ap); // counts args by number of % symbols in fmt, excluding %%; can only handle void* sizes (ie no float/double!)
 mp_obj_t mp_obj_new_fun_bc(mp_obj_t def_args, mp_obj_t def_kw_args, const byte *code, const mp_uint_t *const_table);
 mp_obj_t mp_obj_new_fun_native(mp_obj_t def_args_in, mp_obj_t def_kw_args, const void *fun_data, const mp_uint_t *const_table);
 mp_obj_t mp_obj_new_fun_viper(size_t n_args, void *fun_data, mp_uint_t type_sig);

py/objexcept.c

@@ -114,7 +114,15 @@ void mp_obj_exception_print(const mp_print_t *print, mp_obj_t o_in, mp_print_kin
         } else if (o->args->len == 1) {
             // try to provide a nice OSError error message
             if (o->base.type == &mp_type_OSError && MP_OBJ_IS_SMALL_INT(o->args->items[0])) {
-                const char* msg = mp_errno_to_str(o->args->items[0]);
+                const compressed_string_t* common = mp_common_errno_to_str(o->args->items[0]);
+                const char* msg;
+                if (common != NULL) {
+                    char decompressed[common->length];
+                    decompress(common, decompressed);
+                    msg = decompressed;
+                } else {
+                    msg = mp_errno_to_str(o->args->items[0]);
+                }
                 if (msg[0] != '\0') {
                     mp_printf(print, "[Errno " INT_FMT "] %s", MP_OBJ_SMALL_INT_VALUE(o->args->items[0]), msg);
                     return;
@@ -311,7 +319,7 @@ mp_obj_t mp_obj_new_exception_args(const mp_obj_type_t *exc_type, size_t n_args,
     return exc_type->make_new(exc_type, n_args, 0, args);
 }
 
-mp_obj_t mp_obj_new_exception_msg(const mp_obj_type_t *exc_type, const char *msg) {
+mp_obj_t mp_obj_new_exception_msg(const mp_obj_type_t *exc_type, const compressed_string_t *msg) {
     return mp_obj_new_exception_msg_varg(exc_type, msg);
 }
 
@@ -349,7 +357,7 @@ STATIC void exc_add_strn(void *data, const char *str, size_t len) {
 }
 
 
-mp_obj_t mp_obj_new_exception_msg_varg(const mp_obj_type_t *exc_type, const char *fmt, ...) {
+mp_obj_t mp_obj_new_exception_msg_varg(const mp_obj_type_t *exc_type, const compressed_string_t *fmt, ...) {
     va_list ap;
     va_start(ap, fmt);
     mp_obj_t exception = mp_obj_new_exception_msg_vlist(exc_type, fmt, ap);
@@ -357,7 +365,7 @@ mp_obj_t mp_obj_new_exception_msg_varg(const mp_obj_type_t *exc_type, const char
     return exception;
 }
 
-mp_obj_t mp_obj_new_exception_msg_vlist(const mp_obj_type_t *exc_type, const char *fmt, va_list ap) {
+mp_obj_t mp_obj_new_exception_msg_vlist(const mp_obj_type_t *exc_type, const compressed_string_t *fmt, va_list ap) {
     assert(fmt != NULL);
 
     // Check that the given type is an exception type
@@ -365,7 +373,7 @@ mp_obj_t mp_obj_new_exception_msg_vlist(const mp_obj_type_t *exc_type, const cha
 
     // Try to allocate memory for the message
     mp_obj_str_t *o_str = m_new_obj_maybe(mp_obj_str_t);
-    size_t o_str_alloc = strlen(fmt) + 1;
+    size_t o_str_alloc = fmt->length + 1;
     byte *o_str_buf = m_new_maybe(byte, o_str_alloc);
 
     bool used_emg_buf = false;
@@ -391,15 +399,16 @@ mp_obj_t mp_obj_new_exception_msg_vlist(const mp_obj_type_t *exc_type, const cha
     }
 
     if (o_str_buf == NULL) {
-        // No memory for the string buffer: assume that the fmt string is in ROM
+        // No memory for the string buffer: the string is compressed so don't add it.
-        // and use that data as the data of the string
+        o_str->len = 0;
-        o_str->len = o_str_alloc - 1; // will be equal to strlen(fmt)
+        o_str->data = NULL;
-        o_str->data = (const byte*)fmt;
     } else {
         // We have some memory to format the string
         struct _exc_printer_t exc_pr = {!used_emg_buf, o_str_alloc, 0, o_str_buf};
         mp_print_t print = {&exc_pr, exc_add_strn};
-        mp_vprintf(&print, fmt, ap);
+        char fmt_decompressed[fmt->length];
+        decompress(fmt, fmt_decompressed);
+        mp_vprintf(&print, fmt_decompressed, ap);
         exc_pr.buf[exc_pr.len] = '\0';
         o_str->len = exc_pr.len;
         o_str->data = exc_pr.buf;

py/parsenum.c

@@ -148,11 +148,11 @@ overflow:
 value_error:
     if (MICROPY_ERROR_REPORTING == MICROPY_ERROR_REPORTING_TERSE) {
         mp_obj_t exc = mp_obj_new_exception_msg(&mp_type_ValueError,
-            "invalid syntax for integer");
+            translate("invalid syntax for integer"));
         raise_exc(exc, lex);
     } else if (MICROPY_ERROR_REPORTING == MICROPY_ERROR_REPORTING_NORMAL) {
         mp_obj_t exc = mp_obj_new_exception_msg_varg(&mp_type_ValueError,
-            "invalid syntax for integer with base %d", base);
+            translate("invalid syntax for integer with base %d"), base);
         raise_exc(exc, lex);
     } else {
         vstr_t vstr;
@@ -328,7 +328,7 @@ mp_obj_t mp_parse_num_decimal(const char *str, size_t len, bool allow_imag, bool
     }
 #else
     if (imag || force_complex) {
-        raise_exc(mp_obj_new_exception_msg(&mp_type_ValueError, "complex values not supported"), lex);
+        raise_exc(mp_obj_new_exception_msg(&mp_type_ValueError, translate("complex values not supported")), lex);
     }
 #endif
     else {
@@ -336,9 +336,9 @@ mp_obj_t mp_parse_num_decimal(const char *str, size_t len, bool allow_imag, bool
     }
 
 value_error:
-    raise_exc(mp_obj_new_exception_msg(&mp_type_ValueError, "invalid syntax for number"), lex);
+    raise_exc(mp_obj_new_exception_msg(&mp_type_ValueError, translate("invalid syntax for number")), lex);
 
 #else
-    raise_exc(mp_obj_new_exception_msg(&mp_type_ValueError, "decimal numbers not supported"), lex);
+    raise_exc(mp_obj_new_exception_msg(&mp_type_ValueError, translate("decimal numbers not supported")), lex);
 #endif
 }

py/py.mk

@@ -315,7 +315,7 @@ $(HEADER_BUILD)/qstrdefs.enum.h: $(PY_SRC)/makeqstrdata.py $(HEADER_BUILD)/qstrd
 # the lines in "" and then unwrap after the preprocessor is finished.
 $(HEADER_BUILD)/qstrdefs.generated.h: $(PY_SRC)/makeqstrdata.py $(HEADER_BUILD)/$(TRANSLATION).mo $(HEADER_BUILD)/qstrdefs.preprocessed.h
 	$(STEPECHO) "GEN $@"
-	$(PYTHON3) $(PY_SRC)/makeqstrdata.py --translation $(HEADER_BUILD)/$(TRANSLATION).mo $(HEADER_BUILD)/qstrdefs.preprocessed.h > $@
+	$(PYTHON3) $(PY_SRC)/makeqstrdata.py --compression_filename $(HEADER_BUILD)/compression.generated.h --translation $(HEADER_BUILD)/$(TRANSLATION).mo $(HEADER_BUILD)/qstrdefs.preprocessed.h > $@
 
 $(PY_BUILD)/qstr.o: $(HEADER_BUILD)/qstrdefs.generated.h
 

py/qstr.c

@@ -104,7 +104,7 @@ const qstr_pool_t mp_qstr_const_pool = {
     {
 #ifndef NO_QSTR
 #define QDEF(id, str) str,
-#define TRANSLATION(id, str)
+#define TRANSLATION(id, length, compressed...)
 #include "genhdr/qstrdefs.generated.h"
 #undef TRANSLATION
 #undef QDEF

py/runtime.c

@@ -1539,7 +1539,7 @@ NORETURN void m_malloc_fail(size_t num_bytes) {
         translate("memory allocation failed, allocating %u bytes"), (uint)num_bytes);
 }
 
-NORETURN void mp_raise_msg(const mp_obj_type_t *exc_type, const char *msg) {
+NORETURN void mp_raise_msg(const mp_obj_type_t *exc_type, const compressed_string_t *msg) {
     if (msg == NULL) {
         nlr_raise(mp_obj_new_exception(exc_type));
     } else {
@@ -1547,7 +1547,7 @@ NORETURN void mp_raise_msg(const mp_obj_type_t *exc_type, const char *msg) {
     }
 }
 
-NORETURN void mp_raise_msg_varg(const mp_obj_type_t *exc_type, const char *fmt, ...) {
+NORETURN void mp_raise_msg_varg(const mp_obj_type_t *exc_type, const compressed_string_t *fmt, ...) {
     va_list argptr;
     va_start(argptr,fmt);
     mp_obj_t exception = mp_obj_new_exception_msg_vlist(exc_type, fmt, argptr);
@@ -1555,27 +1555,27 @@ NORETURN void mp_raise_msg_varg(const mp_obj_type_t *exc_type, const char *fmt,
     nlr_raise(exception);
 }
 
-NORETURN void mp_raise_AttributeError(const char *msg) {
+NORETURN void mp_raise_AttributeError(const compressed_string_t *msg) {
     mp_raise_msg(&mp_type_AttributeError, msg);
 }
 
-NORETURN void mp_raise_RuntimeError(const char *msg) {
+NORETURN void mp_raise_RuntimeError(const compressed_string_t *msg) {
     mp_raise_msg(&mp_type_RuntimeError, msg);
 }
 
-NORETURN void mp_raise_ImportError(const char *msg) {
+NORETURN void mp_raise_ImportError(const compressed_string_t *msg) {
     mp_raise_msg(&mp_type_ImportError, msg);
 }
 
-NORETURN void mp_raise_IndexError(const char *msg) {
+NORETURN void mp_raise_IndexError(const compressed_string_t *msg) {
     mp_raise_msg(&mp_type_IndexError, msg);
 }
 
-NORETURN void mp_raise_ValueError(const char *msg) {
+NORETURN void mp_raise_ValueError(const compressed_string_t *msg) {
     mp_raise_msg(&mp_type_ValueError, msg);
 }
 
-NORETURN void mp_raise_ValueError_varg(const char *fmt, ...) {
+NORETURN void mp_raise_ValueError_varg(const compressed_string_t *fmt, ...) {
     va_list argptr;
     va_start(argptr,fmt);
     mp_obj_t exception = mp_obj_new_exception_msg_vlist(&mp_type_ValueError, fmt, argptr);
@@ -1583,11 +1583,11 @@ NORETURN void mp_raise_ValueError_varg(const char *fmt, ...) {
     nlr_raise(exception);
 }
 
-NORETURN void mp_raise_TypeError(const char *msg) {
+NORETURN void mp_raise_TypeError(const compressed_string_t *msg) {
     mp_raise_msg(&mp_type_TypeError, msg);
 }
 
-NORETURN void mp_raise_TypeError_varg(const char *fmt, ...) {
+NORETURN void mp_raise_TypeError_varg(const compressed_string_t *fmt, ...) {
     va_list argptr;
     va_start(argptr,fmt);
     mp_obj_t exception = mp_obj_new_exception_msg_vlist(&mp_type_TypeError, fmt, argptr);
@@ -1600,7 +1600,7 @@ NORETURN void mp_raise_OSError(int errno_) {
 }
 
 
-NORETURN void mp_raise_NotImplementedError(const char *msg) {
+NORETURN void mp_raise_NotImplementedError(const compressed_string_t *msg) {
     mp_raise_msg(&mp_type_NotImplementedError, msg);
 }
 

py/runtime.h

@@ -147,18 +147,18 @@ mp_obj_t mp_import_name(qstr name, mp_obj_t fromlist, mp_obj_t level);
 mp_obj_t mp_import_from(mp_obj_t module, qstr name);
 void mp_import_all(mp_obj_t module);
 
-NORETURN void mp_raise_msg(const mp_obj_type_t *exc_type, const char *msg);
+NORETURN void mp_raise_msg(const mp_obj_type_t *exc_type, const compressed_string_t *msg);
-NORETURN void mp_raise_msg_varg(const mp_obj_type_t *exc_type, const char *fmt, ...);
+NORETURN void mp_raise_msg_varg(const mp_obj_type_t *exc_type, const compressed_string_t *fmt, ...);
-NORETURN void mp_raise_ValueError(const char *msg);
+NORETURN void mp_raise_ValueError(const compressed_string_t *msg);
-NORETURN void mp_raise_ValueError_varg(const char *fmt, ...);
+NORETURN void mp_raise_ValueError_varg(const compressed_string_t *fmt, ...);
-NORETURN void mp_raise_TypeError(const char *msg);
+NORETURN void mp_raise_TypeError(const compressed_string_t *msg);
-NORETURN void mp_raise_TypeError_varg(const char *fmt, ...);
+NORETURN void mp_raise_TypeError_varg(const compressed_string_t *fmt, ...);
-NORETURN void mp_raise_AttributeError(const char *msg);
+NORETURN void mp_raise_AttributeError(const compressed_string_t *msg);
-NORETURN void mp_raise_RuntimeError(const char *msg);
+NORETURN void mp_raise_RuntimeError(const compressed_string_t *msg);
-NORETURN void mp_raise_ImportError(const char *msg);
+NORETURN void mp_raise_ImportError(const compressed_string_t *msg);
-NORETURN void mp_raise_IndexError(const char *msg);
+NORETURN void mp_raise_IndexError(const compressed_string_t *msg);
 NORETURN void mp_raise_OSError(int errno_);
-NORETURN void mp_raise_NotImplementedError(const char *msg);
+NORETURN void mp_raise_NotImplementedError(const compressed_string_t *msg);
 NORETURN void mp_raise_recursion_depth(void);
 
 #if MICROPY_BUILTIN_METHOD_CHECK_SELF_ARG

py/vm.c

@@ -252,7 +252,7 @@ dispatch_loop:
                     if (obj_shared == MP_OBJ_NULL) {
                         local_name_error: {
                             MARK_EXC_IP_SELECTIVE();
-                            mp_obj_t obj = mp_obj_new_exception_msg(&mp_type_NameError, "local variable referenced before assignment");
+                            mp_obj_t obj = mp_obj_new_exception_msg(&mp_type_NameError, translate("local variable referenced before assignment"));
                             RAISE(obj);
                         }
                     }
@@ -1139,7 +1139,7 @@ unwind_return:
                             }
                         }
                         if (obj == MP_OBJ_NULL) {
-                            obj = mp_obj_new_exception_msg(&mp_type_RuntimeError, "no active exception to reraise");
+                            obj = mp_obj_new_exception_msg(&mp_type_RuntimeError, translate("no active exception to reraise"));
                             RAISE(obj);
                         }
                     } else {
@@ -1281,7 +1281,7 @@ yield:
                     } else
 #endif
                 {
-                    mp_obj_t obj = mp_obj_new_exception_msg(&mp_type_NotImplementedError, "byte code not implemented");
+                    mp_obj_t obj = mp_obj_new_exception_msg(&mp_type_NotImplementedError, translate("byte code not implemented"));
                     nlr_pop();
                     fastn[0] = obj;
                     return MP_VM_RETURN_EXCEPTION;

shared-bindings/audiobusio/PDMIn.c

@@ -196,7 +196,7 @@ STATIC mp_obj_t audiobusio_pdmin_obj_record(mp_obj_t self_obj, mp_obj_t destinat
 
     mp_buffer_info_t bufinfo;
     if (MP_OBJ_IS_TYPE(destination, &mp_type_fileio)) {
-        mp_raise_NotImplementedError("");
+        mp_raise_NotImplementedError(translate("Cannot record to a file"));
     } else if (mp_get_buffer(destination, &bufinfo, MP_BUFFER_WRITE)) {
         if (bufinfo.len / mp_binary_get_size('@', bufinfo.typecode, NULL) < length) {
             mp_raise_ValueError(translate("Destination capacity is smaller than destination_length."));

shared-bindings/os/__init__.c

@@ -197,7 +197,7 @@ STATIC mp_obj_t os_urandom(mp_obj_t size_in) {
     mp_int_t size = mp_obj_get_int(size_in);
     uint8_t tmp[size];
     if (!common_hal_os_urandom(tmp, size)) {
-        mp_raise_NotImplementedError("");
+        mp_raise_NotImplementedError(translate("No hardware random available"));
     }
     return mp_obj_new_bytes(tmp, size);
 }

shared-module/audioio/WaveFile.c

@@ -84,7 +84,7 @@ void common_hal_audioio_wavefile_construct(audioio_wavefile_obj_t* self,
     }
     // Get the sample_rate
     self->sample_rate = format.sample_rate;
-    self->len = 512;
+    self->len = 256;
     self->channel_count = format.num_channels;
     self->bits_per_sample = format.bits_per_sample;
 
@@ -114,13 +114,13 @@ void common_hal_audioio_wavefile_construct(audioio_wavefile_obj_t* self,
     self->buffer = m_malloc(self->len, false);
     if (self->buffer == NULL) {
         common_hal_audioio_wavefile_deinit(self);
-        mp_raise_msg(&mp_type_MemoryError, "");
+        mp_raise_msg(&mp_type_MemoryError, translate("Couldn't allocate first buffer"));
     }
 
     self->second_buffer = m_malloc(self->len, false);
     if (self->second_buffer == NULL) {
         common_hal_audioio_wavefile_deinit(self);
-        mp_raise_msg(&mp_type_MemoryError, "");
+        mp_raise_msg(&mp_type_MemoryError, translate("Couldn't allocate second buffer"));
     }
 }
 

supervisor/shared/translate.c

@@ -26,17 +26,61 @@
 
 #include "supervisor/shared/translate.h"
 
+#include <stdbool.h>
+#include <stdint.h>
 #include <string.h>
 
-inline __attribute__((always_inline)) const char* translate(const char* c) {
+#ifndef NO_QSTR
+#include "genhdr/compression.generated.h"
+#endif
+
+char* decompress(const compressed_string_t* compressed, char* decompressed) {
+    uint8_t this_byte = 0;
+    uint8_t this_bit = 7;
+    uint8_t b = compressed->data[this_byte];
+    // Stop one early because the last byte is always NULL.
+    for (uint16_t i = 0; i < compressed->length - 1; i++) {
+        uint32_t bits = 0;
+        uint8_t bit_length = 0;
+        uint32_t max_code = lengths[0];
+        uint32_t searched_length = lengths[0];
+        while (true) {
+            bits <<= 1;
+            if ((0x80 & b) != 0) {
+                bits |= 1;
+            }
+            b <<= 1;
+            bit_length += 1;
+            if (this_bit == 0) {
+                this_bit = 7;
+                this_byte += 1;
+                b = compressed->data[this_byte]; // This may read past the end but its never used.
+            } else {
+                this_bit -= 1;
+            }
+            if (max_code > 0 && bits < max_code) {
+                break;
+            }
+            max_code = (max_code << 1) + lengths[bit_length];
+            searched_length += lengths[bit_length];
+        }
+        decompressed[i] = values[searched_length + bits - max_code];
+    }
+
+    decompressed[compressed->length-1] = '\0';
+    return decompressed;
+}
+
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wreturn-local-addr"
+inline __attribute__((always_inline)) const compressed_string_t* translate(const char* original) {
     #ifndef NO_QSTR
     #define QDEF(id, str)
-    #define TRANSLATION(id, str) if (strcmp(c, id) == 0) { return str; } else
+    #define TRANSLATION(id, len, compressed...) if (strcmp(original, id) == 0) { static compressed_string_t v = {.length = len, .data = compressed}; return &v; } else
     #include "genhdr/qstrdefs.generated.h"
     #undef TRANSLATION
     #undef QDEF
     #endif
-    {
+    return NULL;
-        return "";
-    }
 }
+#pragma GCC diagnostic pop

supervisor/shared/translate.h

@@ -27,6 +27,15 @@
 #ifndef MICROPY_INCLUDED_SUPERVISOR_TRANSLATE_H
 #define MICROPY_INCLUDED_SUPERVISOR_TRANSLATE_H
 
-const char* translate(const char* c);
+#include <stdint.h>
+
+typedef struct {
+    uint16_t length;
+    const uint8_t data[];
+} compressed_string_t;
+
+const compressed_string_t* translate(const char* c);
+
+char* decompress(const compressed_string_t* compressed, char* decompressed);
 
 #endif  // MICROPY_INCLUDED_SUPERVISOR_TRANSLATE_H

tools/huffman

@@ -0,0 +1 @@
+Subproject commit 27b1bba76198a0b343f694a6d680b5293d1c56aa