Hi Troy,
Thanks once more for all the details! I appreciate it.
I just grepped quickly in the SWORD source code (grep -r "upperUTF8" .
| grep -v ".svn") and the method upperUTF8 appears to be only used in
the following places:
./src/keys/versekey.cpp: stringMgr->upperUTF8(abbr, (unsigned
int)(strlen(abbr)*2));
./src/keys/versekey.cpp: stringMgr->upperUTF8(abbr, (unsigned
int)(strlen(abbr)*2));
./utilities/imp2gbs.cpp:
StringMgr::getSystemStringMgr()->upperUTF8(keyBuffer.getRawData(), size-2);
I think neither of those is currently used in Ezra Project, though. At
the moment I do not have the use case to parse verse keys based on any
special Unicode inputs. I am only using the standard English
abbreviations for verse keys and that only happens internally. So, in
this case I may just process the locales.d files directly in node.js /
JavaScript.
Regarding node-sword-interface and the build process for mobile
platforms ... currently I have only tried Android, which works fine. iOS
should technically work as well, but I have not tried that yet. The
boiler plate work to make all that happen smoothly is provided by the
nodejs-mobile <https://code.janeasystems.com/nodejs-mobile> cordova
plugin. That plugin contains build scripts that seemlessly compile any
native node.js addons like node-sword-interface or also the sqlite3
module that I am using.
And since I am now using an API compatible runtime environment both for
Electron/nodejs and Cordova/nodejs-mobile I did not have to add any
additional glue code. One risk I see with this approach is that the guys
who provide nodejs-mobile discontinue their work for some reason. It's
essentially a completely separately maintained fork of nodejs (it has
nothing to do with V8 actually). Originally it is based on the
ChakraCore JavaScript engine of the Microsoft Edge browser. But the
nodejs-mobile guys ported it to Android and iOS ...
Regarding the StringMgr native callback possibility ... yes technically
this is possible with a node native addon like node-sword-interface.
I am using such a functionality for the InstallMgr and search progress
feedbacks already.
So, long story short ... if in the future a usecase comes up to parse
Unicode-based VerseKeys, I will implement a special StringMgr binding as
you suggested. But for now I'll focus on handling the locales.d content
directly in JavaScript / node.js.
I will keep you posted.
Best regards,
Tobias
On 2/6/21 11:59 PM, Troy A. Griffitts wrote:
The data is pulled from the locales.d/ files, but the toUpper logic is
necessary in a number of places in the engine. Two come to mind
immediately:
parsing verse references not sensitive to case
parsing LD module keys not sensitive to case
To be able to get an uppercase representation of any Unicode
character, it takes a pretty hefty dataset of all known human
languages-- that's why we leave it up to an external library. And
yeah, because ICU is so large, that's why I don't compile it into my
binaries in Bishop. Bishop is about 13MB total, which includes ~8MB
of default module data (KJV, SME, StrongsGreek, StrongsHebrew).
That's about 5MB for the app. If I included ICU, it would greatly
increase the size. And both iOS and Android (Swift and Java) already
have facilities for getting the toUpper of a string.
I hope you can steal the few lines from Bishop's native SWORD code
which tells SWORD to call either Java or Swift when toUpperUTF is called.
I am sorry that this might break the nice ability to have exactly the
same code on both iOS and Android (I am surprised that absolutely no
changes were required for you to interface to a native library on both
iOS and Android! cordova required me to provide: Android: Java-jni
layer; iOS: Swift layer. I am jealous.)
If you can think of an alternative, I am happy to listen. We could
provide a better StringMgr default (I think we simply have a latin-1
single byte tranformation table for basically ASCII characters), which
includes an SW_u32 hash which included German characters, but that's
going to limit the languages we support to only the ones we add to our
toUpper hash, and that's not really a dataset I want to maintain.
Open to suggestions,
Troy
On 2/6/21 2:56 PM, Tobias Klein wrote:
Dear Troy,
Thank you for these explanations! I appreciate it!
For Ezra Project on Android, I am at this point simply compiling
node-sword-interface with the Android cross compilers and it works.
However, as I wrote, I have issues for the German Bible book names now.
Is the StringMgr functionality only used to handle the locales.d
files? Or also for some content inside any SWORD modules?
If it is only used for handling the locales.d files then I would
consider handling the Sword locales.d files directly from JavaScript
/ node.js, which already supports Unicode.
I also checked whether I can cross-compile the ICU library and that
worked, but this is a huge binary (I think 20-30 MB) and I would
rather keep the APK size as small as possible.
Best regards,
Tobias
*From: *Troy A. Griffitts <mailto:scr...@crosswire.org>
*Sent: *Sonntag, 31. Januar 2021 18:20
*To: *sword-devel@crosswire.org <mailto:sword-devel@crosswire.org>
*Subject: *Re: [sword-devel] Sword Locales / German Umlaut Issues /
AndroidBuild
Dear Tobias,
My apologies for taking so long to respond to this, but I wanted to
give a thorough answer. See the summary at the end if you don't care
about the details.
So, SWORD has a class StringMgr, which manages strings within SWORD,
and by default SWORD includes a very basic implementation, which
doesn't necessarily know about or support anything beyond what the
basic C string methods support.
I am sure this invokes a sense of horror from you at first, so let me
explain a bit how we properly handle character sets. First, short
background: since we existed well before the Unicode world, we have
multiple locale files for each language, which you will still see in
the locales.d/ folder, each specifying their character encoding, and
most of the time SWORD doesn't need to manipulate characters, so
simply holding data, and passing that data to a display frontend, and
specifying a font which will handle that encoding was enough in the
old world. IMPORTANT: the one place we do need to manipulate
character data is to perform case-insensitive comparisons. We did
this in the past by converting a string to uppercase before
comparison. You'll notice this in the section for Bible book
abbreviation in each locale-- the partial match key must be in a
toupper state.
Today, everything in SWORD prefers Unicode and specifically, encoded
as UTF-8. To support this:
First, we have utility functions within SWORD for working with
Unicode encoded strings, see:
http://crosswire.org/svn/sword/trunk/include/utilstr.h
Specifically:
SWBuf assureValidUTF8(const char *buf);
SW_u32 getUniCharFromUTF8(const unsigned char **buf, bool skipValidation =
false);
SWBuf *getUTF8FromUniChar(SW_u32 uchar, SWBuf *appendTo);
SWBuf utf8ToWChar(const char *buf);
SWBuf wcharToUTF8(const wchar_t *buf);
To wrap this up, by subclassing StringMgr, SWORD supports
implementing character encoding by linking to other libraries, e.g.,
ICU, Qt, etc. to handle full Unicode support. And while the
StringMgr interface allow implementation of many string functions,
upperUTF8 is the only real method the SWORD engine needs to work
completely. Some utilities use the other methods in there, but the
engine, only needs this method.
In summary, on Android, you are likely not linking to ICU when you
build the native SWORD binary-- which I don't do either for Bishop.
The Cordova SWORD plugin uses the SWORD java-jni bindings, which use
the Java VM to implement StringMgr:
https://crosswire.org/svn/sword/trunk/bindings/java-jni/jni/swordstub.cpp
Search for: AndroidStringMgr
And on iOS the Cordova plugin uses the Swift libraries to do the
same. This is done by using the SWORD flatapi call to
org_crosswire_sword_StringMgr_setToUpper to provide a Swift
implementation to uppercase a string.
http://crosswire.org/svn/sword/trunk/bindings/cordova/cordova-plugin-crosswire-sword/src/ios/SWORD.swift
I hope this give you the information you need to get things working
for you. Please don't hesitate to ask if you need help,
Troy
On 1/17/21 11:59 AM, Tobias Klein wrote:
Dear Troy,
I'm playing with an Android Build of Sword and I get issues with the
German Umlauts.
So I have issues with Bible book names like Römer, Könige, etc.
The Umlauts are shown as ?.
I'm configuring the SWORD build with CMake like below (without ICU!)
I remember having similar issues on Linux when building without ICU.
How do you build SWORD for Bishop? Any suggestions?
Best regards,
Tobias
-- Check for working CXX compiler:
/opt/Android/SDK/ndk/r21b/toolchains/llvm/prebuilt/linux-x86_64/bin/clang++
-- Check for working CXX compiler:
/opt/Android/SDK/ndk/r21b/toolchains/llvm/prebuilt/linux-x86_64/bin/clang++
-- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Check for working C compiler:
/opt/Android/SDK/ndk/r21b/toolchains/llvm/prebuilt/linux-x86_64/bin/clang
-- Check for working C compiler:
/opt/Android/SDK/ndk/r21b/toolchains/llvm/prebuilt/linux-x86_64/bin/clang
-- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Configuring your system to build libsword.
-- SWORD Version 1008900000
_______________________________________________
sword-devel mailing list:sword-devel@crosswire.org
http://crosswire.org/mailman/listinfo/sword-devel
Instructions to unsubscribe/change your settings at above page
_______________________________________________
sword-devel mailing list: sword-devel@crosswire.org
http://crosswire.org/mailman/listinfo/sword-devel
Instructions to unsubscribe/change your settings at above page
_______________________________________________
sword-devel mailing list: sword-devel@crosswire.org
http://crosswire.org/mailman/listinfo/sword-devel
Instructions to unsubscribe/change your settings at above page
