Support for translations

[vidya88]

@mpsido commented on Nov 30, 2018

The firmware code contains tags on the strings intended to be displayed to user, that may allow translation in the future.
Shall we work on this ?

[vidya88]

@Dawn-Piper commented on Dec 4, 2018

not now

[vidya88]

@mpsido commented on Dec 21, 2018

TODO: check if it is possible to do a chinese translation

[vidya88]

@mpsido commented on Dec 21, 2018

Check what we can do with this: https://github.com/zst-embedded/STM32-LCD_Font_Generator

[vidya88]

@mpsido commented on Dec 21, 2018

Every text meant to be shown to the user in the code should be wrapped with _() , it ignored by C compiler thanks to gettext.h

[vidya88]

@mpsido commented on Dec 21, 2018

I suspect the transformation of text goes with the scripts in gen/fonts

[vidya88]

@mpsido commented on Dec 25, 2018

fonts.c include font.inc and fontfixed.inc
These two files contain the same letters but displayed a bit differently (which means we have two possible fonts)
fonts.c defines two functions fontCharWidth and fontCharData

These functions are called in : oled.c::oledDrawChar

We have two types of font: FONT_STANDARD and FONT_FIXED (defined in fonts.h)

In trezor-mcu/gen/fonts/ there is a script generate.py which transforms a png containing letters in the right order into fonts.inc files

[vidya88]

@mpsido commented on Dec 25, 2018

Chinese characters map here: http://www.khngai.com/chinese/charmap/tbluni.php?page=5

[vidya88]

@mpsido commented on Dec 26, 2018

Lowest possible resolution for Chinese Characters: https://chinese.stackexchange.com/questions/16669/lowest-pixel-resolution-needed-to-support-chinese

[vidya88]

@mpsido commented on Jan 4

Could not find any chinese character set with a resolution lower than 16*16 pixels, which means that if we manage to make a chinese translations the messages will show big on the screen.

Question left to answer: how many of such characters can we store in the font file before reaching the device memory limit ?

Challenge: Draw characters instead of bitmaps (the prototype code succeeded in drawing characters as bitmaps, but not as actual charactes).

Challenge: Choose an encoding that can be intuitive for the message writer

[vidya88]

@mpsido commented on Jan 4

Check a python code generating files containing the characters we choose as a bitmap:
https://stackoverflow.com/questions/47545359/drawing-multilingual-text-using-pil-and-saving-as-1-bit-and-8-bit-bitmaps

[vidya88]

@mpsido commented on Jan 8

Here are the encoding of some commonly used chinese characters.
你 = e4bda0
好 = e5a5bd
世 = e4b896
界 = e7958c
我 = e68891
不 = e4b88d
知 = e79fa5
道 = e98193
他 = e4bb96
是 = e698af

I have many problems here:
The current characters file is font.inc, it contains an array of 0x7f elements containing the pixel description of every character we use.

There are two challenges:

Font size challenge

the chinese characters require 12 pixels square to show clearly on the screen. which is higher than the FONT_HEIGHT defined in fonts.h

Encoding challenge

The encoding of the characters belongs to a large scale of possible addresses, even if we use a restrained dictionary of characters we need a big array with lot of empty spaces between the characters to be able to "write naturally in the code"
By "Writing naturally" I mean writing the chinese characters as text in the code like : "你好"

The non natural way would mean customize an enumeration of a small number of characters we are gonna use and encode them one by one lets say 0x18: 你 0x19: 好 etc...
Then in the source code instead of writing "你好" we would write "\x18\x19".

We can have a sort of "conversion script" for that but that would be a headache to update text, especially for non developers.

The other solution would be to use a "map" structure, but that does not exist in C (without ++) we will need to code one.

Bitmap vs font challenge

I could find online a python script able to generate bitmaps from font ttc or ttf files:
This is gonna be very helpful and I could already print some characters on the screen.

But the text is printed in another format, basically the pixels are not situated at the same place in the byte array.
With some big differences: one byte represent a row in bitmap, but in character format is is a column.

I did try to write a conversion code with no success so far, but that is not the biggest issue: I worry more about the two previous ones (we could actually rearrange the code to draw bitmaps for chinese text in case we are on "chinese font").

[vidya88]

@mpsido commented on Jan 8

I think I need like 2-3 more hours to code a map like structure plus the code that fetch the character representation in that map if we are using chinese fonts.

[vidya88]

@mpsido commented on Jan 10

Created a pull request #41 that " displays a "你好世界" message on the screen when the "GetVersion" message is received"

[vidya88]

@mpsido commented on Jan 11

To get chinese translation I need :

the exhausive list of all characters you are going to be using
the font you wanna use (in a ttc of ttf file)

[Dawn-Piper]

Brandon?