tinganho / lya

LYA - Localize Your Application

LYA - Localize Your Application

Operating system	Status
Ubuntu
macOS
Windows

Build

Debug Build

Make a debug build by following commands:

git clone --recursive git://github.com/tinganho/l10ns.git
cd l10ns
mkdir Debug
cd Debug
cmake -DCMAKE_BUILD_TYPE=Debug ..
make

Prerequisites

Mac

brew install boost libxml++ gettext

License

Copyright (c) 2014 Tingan Ho Licensed under the Apache 2 license.

lya's Issues

Find files should be variadic

find_files should be variadic instead of implementing multiple signatures and overload.

Add third party libraries in a CMake file with ExternalProject_Add:s

Linking other sub-tree can be made much easier with CMake's ExternalProject_Add function.

See:
https://cmake.org/cmake/help/v3.2/module/ExternalProject.html

Resolve paths should be variadic

Resolve paths should be variadic instead of implementing multiple signatures.

Make text_writer.text private and add a finished function

Too safely guard us from premature finishes, such as getting the text during placeholder writes or saved text cursors that has not been disposed. It would be wise to implement an end function that retries the text and validates that there is no pending placeholder writes and saved text cursors remaining.

UTF8 code point iterator and file streams

In the LYA project, I struggled with optimising the code point iteration for the various token scanner it uses. Right now, I just convert to utf32 string and take the value of the 4 bytes. This is not the optimal way of doing this. Instead we can get the code point directly from decoding a utf8 string. Utf8 is variable length encoded (1-4 bytes), whereas Utf32 is fixed length encoded. We must pay a memory (and perf?) cost to convert it to utf32. Since most programming language source code contains ASCI character, which are 1 byte encoded, we myst pay a very high (near 4 times in memory) cost to convert the whole string.

This link summarizes the operation quite well:
http://www.nubaria.com/en/blog/?p=371

To further, optimize the memory footprint, we should stream the file reading. Right now, we read in the whole file instead of one memory page 4KB.

Refactor so the javascript language has its own server

The JavaScript plugin which supports TypeScript and JavaScript as a language should be refactored so it can pass in an argument which dictates that language.