alberto-balsalm asked:

so i've noticed you've made contributions to pytumblr on github. would you be willing to put in a small pull request for me? just adding a couple of functions for tumblr API end points that aren't implemented in the python package. i use my github for professional stuff so I don't want anyone on github to know that I use tumblr

Yes, although I’d prefer to make the change to my pytumblr2 (post, github link) rather than trying to contribute to the official pytumblr.

I’ve tried the latter, and it wasn’t effective at getting the change onto PyPI and into the hands of users.

The official pytumblr is drastically of date at this point – it has zero NPF support and its model of the API hasn’t been updated with any of the changes since the introduction of NPF. At this point, it’s somewhere between “unusable” and “partially usable if you already know what you’re doing.”

(If you’re wondering about the notifications endpoint, I want to add that to pytumblr2 when I get time for it. The shitty-but-it-works notification-fetching code I actually use in my bot is here, fwiw.)

ajulian asked:

Fyi, frank or Tumblr appears to have a bug. Per this post: https://ajulian.tumblr.com/post/650718036140130304/frankie-baby-how-you-doing-havent-heard-from

Frank's first reply is deleted and each subsequent reply has switched authors.

Screenshot included

Thanks for pointing it out.

This is definitely a tumblr bug rather than a Frank bug.

There have been a lot of recent tumblr bugs related to the rollout of the Beta editor. Frank does the equivalent of posting through the old (non-Beta) web editor, which often exposes these bugs.

I don’t think I’ve seen this exact one before, although it’s reminiscent of the blank answer bug.

——————–

BTW, did you send the original ask on mobile or web? That might help debug it.

——————–

(Technical details below)

——————–

The OP

In the OP, the ask contains a line with no text, like a double line break.

When I fetch the OP as NPF, this appears (correctly) as a block with empty-string text, like this:

The layout for the OP contains a correctly formed ask entry, together with an incorrect rows entry, similar to the blank answer bug.

Nonetheless, the OP renders okay on mobile and desktop dash, albeit without the double newline.

Reblogs

When the OP is represented as a trail entry in NPF, the empty-string block is not present:

However, the original ask layout entry is preserved, marking blocks 0-2 as the ask.

Thus, in reblogs, Frank’s original reply becomes part of the ask.

All subsequent blocks are shifted up by one from their intended position in the layout.

So the author will switch incorrectly, in a manner dependent on whether each reblog has an odd or even number of blocks. This explains @ajulian’s screenshot.

——————–

@cyle

EDIT: I noticed that I can’t actually see the shifted speaker phenomenon on web or iOS. (I do see the OP itself getting mangled, but reblogs are fine.)

So, the shifted speaker thing may be Android-specific?

fatsexybitch asked:

Hey i think the web/mobile display bug is back. I can't submit to help on my phone, but here's an example (since they follow you) nostalgebraist-autoresponder.tumblr . com/post/649634965468889088/nostalgebraist-autoresponder-seyvetch

Yeah, this is definitely some kind of NPF bug … not sure I’ve seen this one before?

The post is https://nostalgebraist-autoresponder.tumblr.com/post/649634965468889088/nostalgebraist-autoresponder-seyvetch

When I try to edit it, I see the full conversation in the legacy editor, but only the “Yep!” (like the left screenshot) in the beta editor.  So, looks like it’s a bug in conversion to/from NPF, rather than a frontend display bug

@cyle if you’re interested

admiral-craymen asked:

What's with all of the blank answers from Frank lately?

cyle:

nostalgebraist:

nostalgebraist:

nostalgebraist:

Could you link me to an example?

I don’t see anything fitting that description in the last few pages, and Frank makes too many posts these days for me to read all of them :)

@admiral-craymen​

Maybe it’s Tumblr? This looks blank on my dash but not when I go to the post. https://nostalgebraist-autoresponder.tumblr.com/post/648044658444959744/thoughts-on-vintage-computers

Oh yeah, I see it now… this does seem like a new tumblr display bug.

The web UI has been going through some weirdness lately.  Yesterday I noticed I could now compose indented lists, even in the non-beta editor, and got briefly excited … until I tried to post one, and found that indented lists still don’t display correctly on the web dash!

My findings so far:

—-

I used the API to get these posts in both NPF and legacy formats. I also did this with some non-bugged asks, and some other test cases.

The bugged posts are malformed in NPF, in a specific way. A typical case looks like:

{‘type’: 'ask’, 'blocks’: [0], …}

{'type’: 'rows’, 'display’: [{'blocks’: [0]}, {'blocks’: [1]}], …}

In NPF answer posts, the layout entry of type “ask” designates some content blocks as part of the ask. All other blocks are assumed to be part of the answer.

Here, block 0 is designated as part of the ask. Then, the second layout entry tries to create a “rows” layout including blocks 0 and 1. But block 0 is in the ask, and block 1 is in the answer.

A guess: the UI thinks that if block 1 is rendered at all, it should render as part of the “rows” group it belongs to. But that group cannot be rendered at all … so it just skips block 1.

The posts look normal in legacy format, which matches the fact that they look okay in the https://username.tumblr.com/post/id view. (Possibly dependent on theme)

I have not been able to reproduce the behavior manually. I have tried the following matrix of cases:

• Ask sent via: mobile, web NPF (new), web legacy
• Response made with: web legacy, API legacy

@cyle you probably know all this but figured I’d tag you

EDIT: the fact that this is even possible seems like a … if not necessarily a flaw, then a weird property of NPF. I definitely remember that when I read the spec, I was surprised at how loosely the ask/answer distinction was enforced.

yeah this is most certainly a bug on our side, we’ll get it fixed

admiral-craymen asked:

What's with all of the blank answers from Frank lately?

nostalgebraist:

nostalgebraist:

Could you link me to an example?

I don’t see anything fitting that description in the last few pages, and Frank makes too many posts these days for me to read all of them :)

@admiral-craymen​

Maybe it’s Tumblr? This looks blank on my dash but not when I go to the post. https://nostalgebraist-autoresponder.tumblr.com/post/648044658444959744/thoughts-on-vintage-computers

Oh yeah, I see it now… this does seem like a new tumblr display bug.

The web UI has been going through some weirdness lately.  Yesterday I noticed I could now compose indented lists, even in the non-beta editor, and got briefly excited … until I tried to post one, and found that indented lists still don’t display correctly on the web dash!

My findings so far:

—-

I used the API to get these posts in both NPF and legacy formats. I also did this with some non-bugged asks, and some other test cases.

The bugged posts are malformed in NPF, in a specific way. A typical case looks like:

{‘type’: 'ask’, 'blocks’: [0], …}

{'type’: 'rows’, 'display’: [{'blocks’: [0]}, {'blocks’: [1]}], …}

In NPF answer posts, the layout entry of type “ask” designates some content blocks as part of the ask. All other blocks are assumed to be part of the answer.

Here, block 0 is designated as part of the ask. Then, the second layout entry tries to create a “rows” layout including blocks 0 and 1. But block 0 is in the ask, and block 1 is in the answer.

A guess: the UI thinks that if block 1 is rendered at all, it should render as part of the “rows” group it belongs to. But that group cannot be rendered at all … so it just skips block 1.

The posts look normal in legacy format, which matches the fact that they look okay in the https://username.tumblr.com/post/id view. (Possibly dependent on theme)

I have not been able to reproduce the behavior manually. I have tried the following matrix of cases:

• Ask sent via: mobile, web NPF (new), web legacy
• Response made with: web legacy, API legacy

@cyle you probably know all this but figured I’d tag you

EDIT: the fact that this is even possible seems like a … if not necessarily a flaw, then a weird property of NPF. I definitely remember that when I read the spec, I was surprised at how loosely the ask/answer distinction was enforced.