Prusa-Firmware/lang/translations.md

# Translations

## Workflow

- Build firmware
  - using `build.sh`
  - using `PF-build.sh` with a parameter `-c 1` to keep lang build files after compiling
- change to `lang` folder
- check if lang scripts being able to run with `config.sh`
  - if you get `Arduino main folder: NG` message change in `config.sh` `export ARDUINO=C:/arduino-1.8.5` to `export ARDUINO=<Path to your Arduino IDE folder>`
    -example: `export ARDUINO=D:/Github/Prusa-Firmware/PF-build-env-1.0.6/windows-64`
- run `lang-build.sh en` to create english `lang_en.tmp`, `lang_en.dat` and `lang_en.bin` files
- change in `fw-build.sh` `IGNORE_MISSING_TEXT=1` to `IGNORE_MISSING_TEXT=0` so it stops with error and generates `not_used<variant>.txt` and `not_tran<variant>.txt`
- run modified `fw-build.sh`
  - `not_tran<variant>.txt` should be reviewed and added as these are potential missing translations
    - copy `not_tran<variant>.txt` as `lang_add.txt` 
	  - check if there are things you don't want to translate or must be modified
	  - als check that the strings do not start with `spaces` as the scripts doesn't handle these well at this moment.
	  - run `lang-add.sh lang_add.txt` to add the missing translations to `lang_en.txt` and `lang_en_??.txt`
  - `not_used<variant>.txt` should only contain messages that aren't used in this variant like MK2.5/S vs MK3/S
- run `fw-clean.sh` to cleanup firmware related files
- delete `not_used<variant>.txt` and `not_tran<variant>.txt`
- run `lang-clean.sh` to cleanup language related files
- run `lang-export.sh all` to create PO files for translation these are stored in `/lang/po` folder
  - Send them to translators and reviewers or
  - copy these to `/lang/po/new` and 
  - translate these with POEdit the newly added messages
    - easiest way is to choose `Validate`in POEdit as it shows you `errors` and the `missing translations` / most likely the newly added at the top.
- The new translated files are expected in `/lang/po/new` folder so store the received files these
- run `lang-import.sh <language code (iso639-1)>` for each newly translated language
  - script improvement to import "all" and other things would be great.
- Double check if something is missing or faulty
  - run `PF-build.sh -v all -n 1 -c 1` to compile for all variants files
  - check if there are still some messages in `not_tran<variant>.txt` that need attention
- After approval
  - run `fw-clean.sh` to cleanup firmware related files
  - run `lang-clean.sh` to cleanup language related files
  - change in `fw-build.sh` back to `IGNORE_MISSING_TEXT=1`
  - build your firmware with `build.sh`, `PF-build.sh` or how you normally do it.
  - Check/Test firmware on printer

## Code / usage
There are 2 modes of operation. If `LANG_MODE==0`, only one language is being used (the default compilation approach from plain Arduino IDE).
The reset of this explanation is devoted to `LANG_MODE==1`:

`language.h`:
```C++
// section .loc_sec (originally .progmem0) will be used for localized translated strings
#define PROGMEM_I2 __attribute__((section(".loc_sec")))
// section .loc_pri (originally .progmem1) will be used for localized strings in english
#define PROGMEM_I1 __attribute__((section(".loc_pri")))
// section .noloc (originally progmem2) will be used for not localized strings in english
#define PROGMEM_N1 __attribute__((section(".noloc")))
#define _I(s) (__extension__({static const char __c[] PROGMEM_I1 = "\xff\xff" s; &__c[0];}))
#define ISTR(s) "\xff\xff" s
#define _i(s) lang_get_translation(_I(s))
#define _T(s) lang_get_translation(s)
```
That explains the macros:
- `_i` expands into `lang_get_translation((__extension__({static const char __c[] PROGMEM_I1 = "\xff\xff" s; &__c[0];})))` . Note the two 0xff's in the beginning of the string. `_i` allows for declaring a string directly in-place of C++ code, no string table is used. The downside of this approach is obvious - the compiler is not able/willing to merge duplicate strings into one.
- `_T` expands into `lang_get_translation(s)` without the two 0xff's at the beginning. Must be used in conjunction with MSG tables in `messages.h`. Allows to declare a string only once and use many times.
- `_N` means not-translated. These strings reside in a different segment of memory.

The two 0xff's are somehow magically replaced by real string ID's where the translations are available (still don't know where).
```C++
const char* lang_get_translation(const char* s){
	if (lang_selected == 0) return s + 2; //primary language selected, return orig. str.
	if (lang_table == 0) return s + 2; //sec. lang table not found, return orig. str.
	uint16_t ui = pgm_read_word(((uint16_t*)s)); //read string id
	if (ui == 0xffff) return s + 2; //translation not found, return orig. str.
	ui = pgm_read_word(((uint16_t*)(((char*)lang_table + 16 + ui*2)))); //read relative offset
	if (pgm_read_byte(((uint8_t*)((char*)lang_table + ui))) == 0) //read first character
		return s + 2;//zero length string == not translated, return orig. str.
	return (const char*)((char*)lang_table + ui); //return calculated pointer
}
```

## Files

### `lang_en.txt`
```
#MSG_CRASH_DET_ONLY_IN_NORMAL c=20 r=4
"Crash detection can\x0abe turned on only in\x0aNormal mode"
```

### `lang_en_*.txt`
```
#MSG_CRASH_DET_ONLY_IN_NORMAL c=20 r=4
"Crash detection can\x0abe turned on only in\x0aNormal mode"
"Crash detekce muze\x0abyt zapnuta pouze v\x0aNormal modu"
```
1. a comment - usually a MSG define with number of characters (c) and rows (r)
2. English text
3. translated text

### `not_tran.txt`
A simple list of strings that are not translated yet.

### `not_used.txt`
A list os strings not currently used in this variant of the firmware or are obsolete.
Example: There are MK2.5 specific messages that aren't used when you compile a MK3 variant and vice versa. So be careful and double check the code if this message is obsolete or just not used due to the chosen variant.

## Scripts

### `config.sh`
- Checks setup and sets auxiliary env vars used in many other scripts. 
- Looks for env var `ARDUINO`. If not found/empty, a default `C:/arduino-1.8.5` is used.
- Sets env var `CONFIG_OK=1` when all good, otherwise sets `CONFIG_OK=0`

### `fw-build.sh`
Joins firmware HEX and language binaries into one file.

### `fw-clean.sh`

### `lang-add.sh`
Adds new messages into the dictionary regardless of whether there have been any older versions.

### `lang-build.sh`
Generates lang_xx.bin (language binary files) for the whole firmware build.

Arguments:
- `$1` : language code (`en`, `cz`, `de`, `es`, `fr`, `it`, `pl`) or `all`
- empty/no arguments defaults to `all`

Input: `lang_en.txt` or `lang_en_xx.txt`

Output: `lang_xx.bin`

Temporary files: `lang_xx.tmp` and `lang_xx.dat`

Description of the process:
The script first runs `lang-check.py $1` and removes empty lines and comments (and non-translated texts) into `lang_$1.tmp`.
The tmp file now contains all translated texts (some of them empty, i.e. "").
The tmp file is then transformed into `lang_$1.dat`, which is a simple dump of all texts together, each terminated with a `\x00`.
Format of the `bin` file:
- 00-01: `A5 5A`
- 02-03: `B4 4B`
- 04-05: 2B size
- 06-07: 2B number of strings
- 08-09: 2B checksum
- 0A-0B: 2B lang code hex data: basically `en` converted into `ne`, i.e. characters swapped. Only `cz` is changed into `sc` (old `cs` ISO code).
- 0C-0D: 2B signature low
- 0E-0F: 2B signature high
- 10-(10 + 2*number of strings): table of string offsets from the beginning of this file
- after the table there are the strings themselves, each terminated with `\x00`

The signature is composed of 2B number of strings and 2B checksum in lang_en.bin. Signature in lang_en.bin is zero.

### `lang-check.sh` and `lang-check.py`
Both do the same, only lang-check.py is newer, i.e. lang-check.sh is not used anymore.
lang-check.py makes a binary comparison between what's in the dictionary and what's in the binary.

### `lang-clean.sh`
Removes all language output files from lang folder. That means deleting:
-  if [ "$1" = "en" ]; then
  rm_if_exists lang_$1.tmp
 else
  rm_if_exists lang_$1.tmp
  rm_if_exists lang_en_$1.tmp
  rm_if_exists lang_en_$1.dif
  rm_if_exists lang_$1.ofs
  rm_if_exists lang_$1.txt
 fi
 rm_if_exists lang_$1_check.dif
 rm_if_exists lang_$1.bin
 rm_if_exists lang_$1.dat
 rm_if_exists lang_$1_1.tmp
 rm_if_exists lang_$1_2.tmp

### `lang-export.sh`
Exports PO (gettext) for external translators.

### `lang-import.sh`
Import from PO.

Arguments:
- `$1` : language code (`en`, `cz`, `de`, `es`, `fr`, `it`, `pl`)
- empty/no arguments quits the script

Input files: `<language code>.po` files like `de.po`, `es.po`, etc.

Input folder: ´/lang/po/new´

Output files:

Output foler: ´/lang/po/new´

Needed improvements to scripts:
 - add `all` argument
 - update `replace in <language> translations` to all known special characters the LCD display with Japanese ROM cannot display
 - move `lang_en_<language code>.txt` to folder `/lang`
 - cleanup `<language code>_filtered.po`, `<language code>_new.po` and `nonasci.txt`

### `progmem.sh`

Examine content of progmem sections (default is progmem1).

Input:
- $OUTDIR/Firmware.ino.elf
- $OUTDIR/sketch/*.o (all object files)

Outputs:
- text.sym     - formatted symbol listing of section '.text'
- $PROGMEM.sym - formatted symbol listing of section '.progmemX'
- $PROGMEM.lss - disassembly listing file
- $PROGMEM.hex - variables - hex
- $PROGMEM.chr - variables - char escape
- $PROGMEM.var - variables - strings
- $PROGMEM.txt - text data only (not used)

Description of process:
- check input files
- remove output files
- list symbol table of section '.text' from output elf file to text.sym (sorted by address)
- calculate start and stop address of section '.$PROGMEM'
- dump $PROGMEM data in hex format, cut disassembly (keep hex data only) into $PROGMEM.lss
- convert $PROGMEM.lss to $PROGMEM.hex:
 - replace empty lines with '|' (variables separated by empty lines)
 - remove address from multiline variables (keep address at first variable line only)
 - remove '<' and '>:', remove whitespace at end of lines
 - remove line-endings, replace separator with '\n' (join hex data lines - each line will contain single variable)
- convert $PROGMEM.hex to $PROGMEM.chr (prepare string data for character check and conversion)
 - replace first space with tab
 - replace second and third space with tab and space
 - replace all remaining spaces with '\x'
 - replace all tabs with spaces
- convert $PROGMEM.chr to $PROGMEM.var (convert data to text) - a set of special characters is escaped here including `\x0a`


### `textaddr.sh`

Compiles `progmem1.var` and `lang_en.txt` files to `textaddr.txt` file (mapping of progmem addresses to text identifiers).

Description of process:
- check if input files exists
- create sorted list of strings from progmem1.var and lang_en.txt
- lines from progmem1.var will contain address (8 chars) and english text
- lines from lang_en.txt will contain line number and english text
- after sort this will generate pairs of lines (line from progmem1 first)
- result of sort is compiled with simple script and stored into file textaddr.txt

Input:
- progmem1.var
- lang_en.txt

Output:
- textaddr.txt



update_lang.sh