eric99

23 minutes ago, Mike P said:

I just seasrched for 

notes tagged with $current updated in last 5 days

AND

notes tagged with $current about buying

and got the notes I expected. So I think tagging the notes you want it to search may be possible.

 

 

have you already tried to set an AI search sandbox first by selecting the tag in the left side bar and then search for a specific topic? Do the 8 notes belong to the tag then (if they are shown)?

2 hours ago, BrooklynBen said:

Is there a way to tell it what 8 posts to prioritize? I have a couple major summary notes that link to hundreds of other notes, and those would be the most useful to train it on. 

This would be possible if you could constrain the AI search to a specific tag or notebook, specified in the search string or selected in the side bar before searching.

1 hour ago, Paul A. said:

I don't have specific reproducible steps, but my experience with data loss in recent months involved trying to open a pdf embedded into an older note and having Evernote occasionally "eat" the pdf. I have also seen some posts of people losing text added to a new (or newly edited) note (while the app continued to claim "All changes saved" in the lower-right). Hopefully both of these issues are fixed. 🙏 

Thanks, but my question was more directed to Federico. Now that they've found the problem, maybe they can describe more precisely the circumstances under which it went wrong?

49 minutes ago, janndk said:

Just a simple test note, https://www.evernote.com/shard/s56/sh/71b56831-4b18-7a5e-ea93-0c76a892087e/ql3JonBmZhiqssOXo1kBPHzz0KY2vwuszLUKi2rWDfOJ-HQ4muj8wdpsiA

I exported your note to PDF with numbering intact. I don't see the problem ? Which pdf viewer do you use?

edit: Sorry, I was mislead by the message that all numbers were zero, this is not the case (anymore) but your issue is that the last number is wrong (3 instead of 4). I can reproduce that too. So there is still a problem with numbering...

37 minutes ago, johnm243 said:

Thanks for the mention of Directory Opus @eric99: my renewal for Copernic is coming up and Directory Opus sounds like a better full-text search app!

FYI, I will be testing Joplin Cloud soon to see if the test notes I have exported from EN Legacy to Joplin will sync properly to the Joplin App! If all works fine, I will export all from Legacy, test  again ... and hopefully break the cord!

It was CalS' idea for Directory Opus, he's the expert...

2 hours ago, Federico Simionato said:

This is a big one: except for the fixes reported below, this release addresses what’s likely the most common cause of missing content.

Thanks, glad to hear this 👍

In what situations was it possible to lose data?

2 hours ago, eric99 said:

I have written my own Enex to HTML converter and see no problems with pdf or any other attachment  (so far).

I exported all my notes now with the latest EN and I can convert all of them with my own enex to HTML converter without any problem.

So the import problem could be caused by the import software. On Github I've seen several crappy enex converters that use regular expressions instead of normal XML parsers to process the enex file. This makes the converter very fragile and sensitive to insignificant differences in the enex file that are otherwise ignored by XML parsers. I also noticed that these implementations are forked into several projects, so more importers can be affected by the same bug.

6 hours ago, RobertJLee said:

Word of Caution

Generating an ENEX file with latest version results in an import file not containing pdfs included within the various notes. Only file names.An absolute nightmare.

Also tried importing enex into Joplin same result

Bear imports associated PDFS so.....EN toENEX ......enex to Bear......export from bear to RTF with Attachments.......import to apple notes

I have written my own Enex to HTML converter and see no problems with pdf or any other attachment  (so far).

2 hours ago, Jon/t said:

Interesting. Or maybe restrict a certain notebook(s) from ever hitting the Ai.

Yes, but my question was even simpler than that: will the 8 notes for the AI always be limited to the specified notebook or tags in the query string (or possibly set via the side bar)? That would help a lot to gradually explore and build trust...

On 9/18/2023 at 2:32 AM, CalS said:

@idoc  Not sure you want to add another product, but... 

When I left EN I purchased Directory Opus, a File Explorer on steroids plus piece of software.  I use it as a gateway to the repository I created from my EN notes.  It even kind of looks like EN with a folder and saved search left panel, a file list middle panel, and a display right panel.

I exported 58k notes to HTML.  I've added another 6k files since then.  Searches in DO can return files and folders.  You can segregate or hide types as you see fit.  You can create saved searches as well.  Depending upon the content of you faxes you may be able to create a search to do the job.  Assuming you index the folders containing these files searches are pretty much as you type.

Or this could be a worse change for your processes.   🤷‍♂️

Can Document Opus search pdf and word documents and is textual search fast?

Can the AI search be limited to a specific notebook so that sensitive notes are never sent to the AI server?

5 hours ago, PinkElephant said:

Forget the updated field, and modify the created field only when you enter information not critical for you. When you for example export to ENEX, and import again, this field is overwritten with the import date (yes, as well on legacy). So better not to entrust it with any vital information.

Hmm,  not sure about that... have you tested this recently?

edit: I just tested this on my Windows client: I exported a note to Enex and re-imported it, the update and creation time did not change, the imported note is an exact clone, except for the unique note-id of course

1 hour ago, prashantjain25 said:

 The culprits here are the commercial OS vendors exerting undue pressure, and the suffocating grasp of patents that stifle Linux's growth.

Don't blame the OS vendors, the real culprit is the Linux community itself: there are over 600 active Linux distributions and counting 😵. To the outside world there doesn't seem to be such a thing as one 'Linux OS'

20 hours ago, PinkElephant said:

As I understand it, the „classical“ search makes a preselection based on the search parameters. This is then presented to the AI filter, which returns a refined and maybe better sorted final search result.

When I tested it, it gave the results back with a „term 1 OR term 2 OR term 3“ search string - which is not what I had entered. So it seems AI took the initial search parameters, and refined it into a sort of Boolean request.

Just an observation, what it really does will only show after some extensive testing - and if the results are really superior to a classical search.

Have you tested with natural language queries to see which query it is translated into by the AI? Typos and the use of synonyms are also an interesting test.

20 hours ago, Jon/t said:

No. It won't have all-access to your data. They're planning to explain how it works when its rolled out  but there's no kind of pipeline into your notes, its much more selective.

If it doesn't have all access to the data, is it more processing the query string then (natural language, synonyms etc) ?

1 hour ago, gazumped said:

I do suspect that if this were ever possible,  it would not be backwards-compatible.  Evernote maybe would have to finally end the Legacy lifeline to provide the access.

I think it will be backwards compatible since it's ENEX XML format. New XML tags for tasks or other new stuff will be ignored by the legacy enex readers. I already tested that by adding a note-id and notebook name. Legacy importers still work fine, they just ignore these two new attributes...

And to make a consistent and complete backup of it, please add the note id and notebook to the ENEX file.

8 minutes ago, s2sailor said:

Thanks.  It is mostly a guess on my part but it makes sense, at least to me.  I'm all for improving what we have.  Instead of tinkering with the backend note history, I would prefer a local, desktop user-definable backup.  RTE updates the server copy.  How about an option where, on a successful update of the server copy,  it would then trigger an update to a local database copy, so in a sense, we would have a local, definable history and the system history we currently have?  One can wish and dream 🙂.

So you mean the note export functionality 😏 You can make as many and as frequently as you like...

15 minutes ago, AlbertR said:

• at least an indicator that note needs to be synced

Fully agree, so not the misleading optimistic message in the bottom right corner that everything has been synced when it's not... (in desktop clients)

2 hours ago, gazumped said:

 Erratic behaviour is often just the cost of doing business via sketchy electronic devices over long distances.

Which is of course the favorite excuse for developers who don't know about concurrent programming techniques to handle race conditions properly 😏

I don't think it's the nbsp char, I inserted it in a title and the title still exports properly.

I then searched the content of the whole local database for  a note which I know is converted to 'Untitled' by enex, checked the title with notepad++ hex viewer but can't find anything special. It's a challenge to reverse engineer the database structure from a bunch of files, maybe I missed it

On 9/5/2023 at 10:11 PM, eric99 said:

Yeah, I built in some extra checks in my own HTML export tool and it discovered 28 Untitled Notes in my latest enex export

I'll try to find out which hidden characters are the culprit and how they end up in the title

Perhaps the odd character is nbsp, which sometimes appears in enex files, making life harder for xml parsers and maybe for the enex generator as well when inside a title? I don't think it is valid enex character according the dtd file.  I look further...

Yeah, I built in some extra checks in my own HTML export tool and it discovered 28 Untitled Notes in my latest enex export

I'll try to find out which hidden characters are the culprit and how they end up in the title

Just FYI, this is the offical ENML doc: https://dev.evernote.com/doc/articles/enml.php#added

1 hour ago, boardtc said:

@eric99, excellent debugging

1. Individual export produces Untitled note.md in enex file
2. Duplicate of note produces correct title.md

I can do that for these 2 notebooks but I would prefer not to have to check my other 30+ by hand, some with 100s of notes. Still not the end of the world with this work around and a bit of grafting. Many thanks.

Thanks for testing!  I suggested this to diagnose the problem but I'm glad that it helps you with your export 🙂

What do we learn from this?  The most probable conclusion is what john/T already suggested: the title contains hidden characters which were removed in the duplicated note. To prove that theory you could manually remove the title in the original note and type it again. Then export to check whether the exported title is still empty.

Maybe this problem should be discussed with @Federico Simionato before destroying the traces ?