zim-wiki team mailing list archive

[Jaap Karssenberg <jaap.karssenberg@xxxxxxxxx>]

Hi Mario,

I did some testing myself last night, will write up the full log in the dev
wiki later but here is the summary:

Generated 3 random notebooks:
1/ 16_000+ files over 600+ folders, total size on disk 200Mb (many small
files)
2/ 16_000+ files over 600+ folders, total size on disk 3Gb (many larger
files)
3/ the same as #1 but with a single ".txt" attachment of 1Gb

Indexing with "zim --index" takes 3-4 minutes for #1 & #3, but over 60
minutes (!) for #2

Once the indexing is complete I open the notebook with zim. For all 3
cases, the UI responds within seconds. The background index check runs for
1-2 minutes but does not interfere with using the application already.

Conclusions:
- Large text attachments are handled (ignored) correctly by the indexer
- Indexing cache allows fast opening of a very large notebook when the
index is up to date
- Indexing speed scales with the amount of text that needs to be read and
parsed. Parsing is not super fast and might be improved, but I consider the
3Gb notebook an extreme scenario; you can spent a life time typing in 3Gb
of text notes ;)

Further thoughts:
- When testing a new version (or e.g. switching between master and develop
branches) you may need to run an index update first for fare comparison,
sometimes an update is needed when the database scheme changed
- The index uses the "mtime" time stamps on files and folders. If your
notebook needs index updates every time you open it, and you have a
non-standard file system, it might be the file system does not represent
mtimes correctly.
- By default the index is in the user cache folder (~/.cache on unix) so
using the same notebook from multiple systems may result in indexing the
same changes multiple times. You can choose to store the index in the
notebook itself (see manual for the "shared" notebook property), however
this may cause trouble if you try to access from both systems at the same
time.

Hope this helps to clarify what to expect,

Jaap


On Tue, Jun 29, 2021 at 12:02 PM Mario Bezzi <
subscriptions.mario.bezzi@xxxxxxxxx> wrote:

> Hi Jaap, thank you for all your help on this.
>
> I just tested the development branch which seems to include your mentioned
> fix. Sadly zim startup took approximately the same time as before.
> Maybe my use case is abnormal? If so I don't want you to spend lots of
> time on this.
>
> Below the output of ./zim.py -V -D. All the time is spent in
> BackgroundCheck. Is this what you would expect? May we (I?) add debug
> messages to track the time spent indexing large files?
>
> Thank you,
> mario
>
> ---------------------
>
> mario_bezzi@XPS-15-9560:~/Downloads/zim-desktop-wiki-develop$ ./zim.py -V
> -D
> DEBUG: Loading config from: <ConfigFile:
> /home/mario_bezzi/.config/zim/preferences.conf>
> DEBUG: New extendable: <zim.plugins.InsertedObjectTypeMap object at
> 0x7f99b7a5eaf0>
> DEBUG: Loading plugin: pageindex
> DEBUG: Loading plugin: pathbar
> DEBUG: Loading plugin: insertsymbol
> DEBUG: Loading plugin: printtobrowser
> DEBUG: Loading plugin: versioncontrol
> DEBUG: Loading plugin: attachmentbrowser
> DEBUG: Loading plugin: bookmarksbar
> DEBUG: Loading plugin: journal
> INFO: This is zim 0.73.5
> DEBUG: Python version is sys.version_info(major=3, minor=8, micro=5,
> releaselevel='final', serial=0)
> DEBUG: Platform is posix
> DEBUG: Running from a source dir:
> /home/mario_bezzi/Downloads/zim-desktop-wiki-develop
> DEBUG: Set XDG_DATA_HOME to /home/mario_bezzi/.local/share
> DEBUG: Set XDG_DATA_DIRS to [<Dir: /usr/share/ubuntu>, <Dir:
> /usr/local/share>, <Dir: /usr/share>, <Dir: /var/lib/snapd/desktop>]
> DEBUG: Set XDG_CONFIG_HOME to /home/mario_bezzi/.config
> DEBUG: Set XDG_CONFIG_DIRS to [<Dir: /etc/xdg/xdg-ubuntu>, <Dir: /etc/xdg>]
> DEBUG: Set XDG_CACHE_HOME to /home/mario_bezzi/.cache
> DEBUG: Connecting to /run/user/1000/zim-0.73.5-43e7ec27
> DEBUG: Got error in dispatch: No such file or directory
> DEBUG: Starting primary process
> DEBUG: Start listening on: /run/user/1000/zim-0.73.5-43e7ec27
> DEBUG: Loading config from: <zim.notebook.info.VirtualFile object at
> 0x7f99b5719040>
> DEBUG: Loading config from:
> /home/mario_bezzi/Dropbox/Documents/Wikis/notebook.zim
> DEBUG: Loading config from:
> /home/mario_bezzi/Dropbox/Documents/Wikis/notebook.zim
> DEBUG: Loading config from:
> /home/mario_bezzi/.cache/zim/notebook-home_mario_bezzi_Dropbox_Documents_Wikis/state.conf
> DEBUG: New extendable: <Notebook: Wikis>
> DEBUG: Load extension: <class
> 'zim.plugins.journal.JournalNotebookExtension'>
> DEBUG: Load extension: <class
> 'zim.plugins.versioncontrol.VersionControlNotebookExtension'>
> INFO: No VCS detected
> DEBUG: Loading config from: <ConfigFile:
> /home/mario_bezzi/.config/zim/style.conf>
> DEBUG: Autosave interval: 15 - use threads: True
> DEBUG: Loading config from: <ConfigFile:
> /home/mario_bezzi/.config/zim/customtools/command shell-usercreated.desktop>
> INFO: Page changed on disk: Home:ZABEXPRF:ZX Manuals
> INFO: Open page: Home:ZABEXPRF:ZX Manuals (Home:ZABEXPRF:ZX Manuals)
> DEBUG: New extendable: <notebookview.NotebookView ZABEXPRF at
> 0x7f99b56bee00 (zim+gui+notebookview+NotebookView at 0x1bbbb60)>
> DEBUG: Load extension: <class
> 'zim.plugins.attachmentbrowser.AttachmentBrowserWindowExtension'>
> DEBUG: Action: toggle_panes(False)
> DEBUG: Action: toggle_panes(True)
> DEBUG: Load extension: <class
> 'zim.plugins.insertsymbol.InsertSymbolPageViewExtension'>
> DEBUG: Load extension: <class
> 'zim.plugins.journal.JournalNotebookViewExtension'>
> DEBUG: Load extension: <class
> 'zim.plugins.pageindex.PageIndexNotebookViewExtension'>
> DEBUG: Load extension: <class
> 'zim.plugins.printtobrowser.PrintToBrowserPageViewExtension'>
> DEBUG: New extendable: <mainwindow.MainWindow object at 0x7f99b56b6300
> (zim+gui+mainwindow+MainWindow at 0x1b9a270)>
> DEBUG: Load extension: <class
> 'zim.plugins.bookmarksbar.BookmarksBarMainWindowExtension'>
> DEBUG: Load extension: <class
> 'zim.plugins.pathbar.PathBarMainWindowExtension'>
> DEBUG: Load extension: <class
> 'zim.plugins.versioncontrol.VersionControlMainWindowExtension'>
> DEBUG: Accelmap: /home/mario_bezzi/.config/zim/accelmap
> DEBUG: Add window: MainWindow
> DEBUG: BackgroundCheck started
> DEBUG: BackgroundCheck finished
>
>
>
> On 6/26/21 7:55 PM, Jaap Karssenberg wrote:
>
> Hi Mario,
>
> Just pushed a fix (29bdea) that improves how we check whether files are a
> zim page or not. Now max 50 characters are being read at the start of the
> file when indexing. If your large files are not "line based" (thus
> resulting in a very long read when trying to read the first line) this
> should fix the issue.
>
> Regards,
>
> Jaap
>
>
> On Sat, Apr 24, 2021 at 10:06 AM Mario Bezzi <
> subscriptions.mario.bezzi@xxxxxxxxx> wrote:
>
>> Hi Jaap, thank you for your help on this.
>>
>> To give you some more details: Of the 3000+ files which size sums up to
>> 2GB, the top 500 account for 1.6GB. Among these the average size is 3.5MB,
>> and each of the top three is in the 250MB range.
>>
>> Please let me know if there is anything I can do to help testing your fix,
>> mario
>>
>> On 4/23/21 2:55 PM, Jaap Karssenberg wrote:
>>
>> Hi Mario,
>>
>> That is not the result I hoped for :(   I will need to generate some
>> random large text files to test & debug on my end.
>>
>> Regards,
>>
>> Jaap
>>
>>
>> On Fri, Apr 23, 2021 at 12:59 PM Mario Bezzi <
>> subscriptions.mario.bezzi@xxxxxxxxx> wrote:
>>
>>> I think I submitted my request circa 2014 under the previous bug
>>> tracking system - was it hosted by Ubuntu-one? - but yes, the idea is
>>> similar.
>>>
>>> I just downloaded the development version, extracted it into a temporary
>>> folder, and ran it via the ./zim.py command.
>>>
>>> Indexing took some 15 minutes. Below a snapshot of what top was saying
>>> about the execution.
>>>
>>> top - 12:45:28 up 3 days, 16:12,  1 user,  load average: 1.87, 1.92, 2.48
>>> Tasks: 356 total,   3 running, 353 sleeping,   0 stopped,   0 zombie
>>> %Cpu(s): 13.0 us,  5.4 sy,  0.0 ni, 81.6 id,  0.0 wa,  0.0 hi,  0.0 si,
>>> 0.0 st
>>> MiB Mem :  31658.1 total,    320.9 free,  19312.0 used,  12025.3
>>> buff/cache
>>> MiB Swap:    976.0 total,      0.0 free,    976.0 used.  10085.6 avail
>>> Mem
>>>
>>>     PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+
>>> COMMAND
>>>
>>>  159310 mario_b+  20   0  771220  80184  43420 R 100.0   0.2  *14:42.13
>>> zim.py*
>>>
>>> Please let me know if there is more I can do.
>>>
>>> Thank you,
>>> mario
>>>
>>> On 4/23/21 11:25 AM, Jaap Karssenberg wrote:
>>>
>>> Yes that explains, those large files will have a big impact on the
>>> indexer.
>>>
>>> You are referring to this issue: Make indexer ignore text files that
>>> are not zim pages · Issue #907 · zim-desktop-wiki/zim-desktop-wiki
>>> (github.com)
>>> <https://github.com/zim-desktop-wiki/zim-desktop-wiki/issues/907> which
>>> is fixed in the development branch and will be in the next release.
>>>
>>> With that fix the indexer will read the first line of each file to
>>> decide whether it is a zim file or not, and if not it will not try to
>>> access the contents.
>>>
>>> Would be great if you have a chance to test the development branch and
>>> see whether it works in practice for your case !
>>>
>>> -- Jaap
>>>
>>>
>>> On Thu, Apr 22, 2021 at 7:32 PM Mario Bezzi <
>>> subscriptions.mario.bezzi@xxxxxxxxx> wrote:
>>>
>>>> The folder contains 3118 ".txt" files, for a total of 2GB of data. Some
>>>> large txt files are attachments. A long time ago I submitted a request to
>>>> avoid indexing these. Not sure it has been fulfilled though.
>>>>
>>>> Thank you,
>>>> mario
>>>>
>>>> On 4/8/21 7:32 PM, Jaap Karssenberg wrote:
>>>>
>>>> Can you indicate how big your notebook folder is? Either an extreme
>>>> case, or some bug making it take much longer than needed.
>>>>
>>>> Op do 8 apr. 2021 15:59 schreef Mario Bezzi <
>>>> subscriptions.mario.bezzi@xxxxxxxxx>:
>>>>
>>>>> Thanks Jaap, I was not aware of this.
>>>>>
>>>>> To give you an idea, I just restarted Zim, and indexing kept a
>>>>> processor 100% busy for 13 minutes to come to an end.  It was nice if this
>>>>> could be avoided.
>>>>>
>>>>> Thank you,
>>>>> mario
>>>>>
>>>>> On 4/8/21 10:06 AM, Jaap Karssenberg wrote:
>>>>>
>>>>> The indexing is not used for searching alone, it is also needed to
>>>>> e.g. present the page tree in the side pane and to track links
>>>>>
>>>>> Op do 8 apr. 2021 09:34 schreef Mario Bezzi <
>>>>> subscriptions.mario.bezzi@xxxxxxxxx>:
>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> I may be the only one, but with my quite large notebooks I do find
>>>>>> the
>>>>>> search function impractical, and for this reason I never use it.
>>>>>> Still,
>>>>>> when it starts, Zim goes crazy for a long time indexing, and I came
>>>>>> to
>>>>>> the conclusion that this is normal.
>>>>>>
>>>>>> If this is the case, I would like to file a requirement to add the
>>>>>> ability to make indexing optional.
>>>>>>
>>>>>> Thank you,
>>>>>> mario
>>>>>>
>>>>>> _______________________________________________
>>>>>> Mailing list: https://launchpad.net/~zim-wiki
>>>>>> Post to     : zim-wiki@xxxxxxxxxxxxxxxxxxx
>>>>>> Unsubscribe : https://launchpad.net/~zim-wiki
>>>>>> More help   : https://help.launchpad.net/ListHelp
>>>>>>
>>>>>
>>>>>
>>>>
>>>
>>
>