Request for hint/help on kiwix-js #1
Comments
|
@cptolemy Hi, you've posted this on a very old personal clone of Kiwix JS that is no longer being used or updated. I now commit all new code to https://github.com/kiwix/kiwix-js, which is actively being developed, and to https://github.com/kiwix/kiwix-js-windows/, which is a PWA and UWP port of Kiwix JS. You should be aware that a lot of changes have happened since sharun-s was last involved in this project, including a change of ZIM format (to use ZSTD compression and webp images), which means that unless he has been updating his code-base, then new ZIMs can no longer be read by his old codebase. There is a further change upcoming too (ZIMs with no namespace). We have recently implemented all these changes in Kiwix JS and Kiwix JS Windows. I have developed Kiwix JS PWA version to use the File System Access API. In Chromium browsers, this nearly gets you what you want, but the user does have to pick a file or a folder first time, and thereafter the user simply has to click a prompt "Allow access to files?". It's a near-native experience. I understand that the API will soon allow installed PWAs to access the FS without further permission prompts (after picking an archive the first time). If you want a packaged ZIM "app", e.g. WikiMed, that automatically loads a packaged archive with no prompts, we have those too, using either the NWJS framework of the Electron framework. For Windows 10, we also have a packaged version using the UWP framework (Store app, but it can be side-loaded). I have only compiled Electron and NWJS for Windows, but of course they could be compiled for Mac and Linux too. Here are some links:
There is a thread on Kiwix JS where another dev has developed a way of accessing ZIM files over https, actually developing sharun-s's work, but it was a bit slow to load and involved a helper index, I believe. I'll ping you on that thread so you can have a read through. To answer your specific queries, Kiwix JS does not read the whole file into memory, that would be impossible of course with the large multi-GB files we often have. It use the File API and its methods to access specific offsets from the start of the file or file-set. Sharun-s's idea was to emulate the same over https using XHR range requests. It works, but is very slow, and depends on the server's capabilities. I believe he wanted to do it using the file:// protocol, to avoid having a server and network issues, but file access is now basically deprecated in all browsers except the obsolete ones: IE11 and Edge Legacy. I think if you try to use PHP, you'll end up with similar network issues that have made all these solutions very slow: binary search involves thousands of loookups as it hops from one dirEntry to another across the index. We have introduced a low-level block cache with a Least Recently Used ejection strategy, which helps a lot, but it's still an issue over networks. Electron and NWJS effectively do exactly want you want without any further back-end: they hack the File object and add a path property to it. This means the code only has to know the relative path to a ZIM file, e.g. a packaged WikiMed or Wikivoyage, and it can be accessed with no user interaction whatsoever. But of course they are quite heavy frameworks with a large overhead (though its small compared to the ZIM file sizes). For the lightest approach I'd recommend the PWA version, because it feels native in Chromium browsers, especially if installed, is very lightweight, updates itself and works offline if no network access is available. However, the user must pick the archive first time, which can be a problem if you have naive users. |
|
Good evening Jaifroid,
I am very sorry for not replying earlier.
Thank you very much for your kindness, help and all the resources you
pointed out - they were exactly what I was looking for.
It's a wonderful project. :)
Best regards
cptolemy
Jaifroid <[email protected]> escreveu no dia domingo, 7/02/2021 à(s)
20:51:
… @cptolemy <https://github.com/cptolemy> Hi, you've posted this on a very
old personal clone of Kiwix JS that is no longer being used or updated. I
now commit all new code to https://github.com/kiwix/kiwix-js, which is
actively being developed, and to
https://github.com/kiwix/kiwix-js-windows/, which is a PWA and UWP port
of Kiwix JS.
You should be aware that a lot of changes have happened since sharun-s was
last involved in this project, including a change of ZIM format (to use
ZSTD compression and webp images), which means that unless he has been
updating his code-base, then new ZIMs can no longer be read by his old
codebase. There is a further change upcoming too (ZIMs with no namespace).
We have recently implemented all these changes in Kiwix JS and Kiwix JS
Windows.
I have developed Kiwix JS PWA version to use the File System Access API
<https://web.dev/file-system-access/>. In Chromium browsers, this nearly
gets you what you want, but the user does have to pick a file or a folder
first time, and thereafter the user simply has to click a prompt "Allow
access to files?". It's a near-native experience. I understand that the API
will soon allow installed PWAs to access the FS without further permission
prompts (after picking an archive the first time).
If you want a packaged ZIM "app", e.g. WikiMed, that automatically loads a
packaged archive with no prompts, we have those too, using either the NWJS
framework of the Electron framework. For Windows 10, we also have a
packaged version using the UWP framework (Store app, but it can be
side-loaded). I have only compiled Electron and NWJS for Windows, but of
course they could be compiled for Mac and Linux too.
Here are some links:
- PWA version, very lightweight (if using Chrome or new Edge, it can
be installed; in Firefox you have to bookmark it rather than install it. It
caches its own code and works offline): *https://pwa.kiwix.org/
<https://pwa.kiwix.org/>*
- WikiMed packaged app using Electron, Win 7/8/10 (no file picking
required): *https://kiwix.github.io/kiwix-js-windows/wikimed-electron.html
<https://kiwix.github.io/kiwix-js-windows/wikimed-electron.html>*
- Wikimed UWP (Win10 only): *https://kiwix.github.io/kiwix-js-windows/wikimed-uwp.html
<https://kiwix.github.io/kiwix-js-windows/wikimed-uwp.html>*
- NWJS version for WinXP/Vista/Win7/8/10: *https://kiwix.github.io/kiwix-js-windows/kiwix-js-nwjs.html
<https://kiwix.github.io/kiwix-js-windows/kiwix-js-nwjs.html>*
There is a thread on Kiwix JS where another dev has developed a way of
accessing ZIM files over https, actually developing sharun-s's work, but it
was a bit slow to load and involved a helper index, I believe. I'll ping
you on that thread so you can have a read through.
To answer your specific queries, Kiwix JS does *not* read the whole file
into memory, that would be impossible of course with the large multi-GB
files we often have. It use the File API and its methods to access specific
offsets from the start of the file or file-set. Sharun-s's idea was to
emulate the same over https using XHR range requests. It works, but is very
slow, and depends on the server's capabilities. I believe he wanted to do
it using the file:// protocol, to avoid having a server and network issues,
but file access is now basically deprecated in all browsers except the
obsolete ones: IE11 and Edge Legacy.
I think if you try to use PHP, you'll end up with similar network issues
that have made all these solutions very slow: binary search involves
thousands of loookups as it hops from one dirEntry to another across the
index. We have introduced a low-level block cache with a Least Recently
Used ejection strategy, which helps a lot, but it's still an issue over
networks.
Electron and NWJS effectively do exactly want you want without any further
back-end: they hack the File object and add a path property to it. This
means the code only has to know the relative path to a ZIM file, e.g. a
packaged WikiMed or Wikivoyage, and it can be accessed with no user
interaction whatsoever. But of course they are quite heavy frameworks with
a large overhead (though its small compared to the ZIM file sizes). For the
lightest approach I'd recommend the PWA version, because it feels native in
Chromium browsers, especially if installed, is very lightweight, updates
itself and works offline if no network access is available. However, the
user must pick the archive first time, which can be a problem if you have
naive users.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/ADXFADXJSCA5KUHTC3EDPITS534L5ANCNFSM4XHWBLRA>
.
--
*Rui Miguel FernandesPorto - Portugal*
|
Good afternoon mossroy
How are you today?
I am contacting you, since sharun-s thought it would be best instead of trying to modify her code of kiwix-html5. I sent this exact request for help to mossroy by email.
Basically, I am trying to automatically read a default zim file in kiwix-js. I understand the browser doesn't give us direct permission, but I was thinking in opening and reading the file via a jquery/ajax call to a php script that would return the contents. But I don't know the exact point where the file is read, if it's read fully, or if by making the change I can simplify the reading by skipping the first html template.
Could you help me please?
sharun-s script is quite good, but to place another zim file proven to be hard.
Thank you so much.
Best regards, and stay safe
CPtolemy
The text was updated successfully, but these errors were encountered: