Re: Serious feedback and questions about the future of pgAdmin.

Hi

On Sun, Jun 11, 2017 at 6:42 PM, Jack Royal-Gordon <jackrg(at)pobox(dot)com> wrote:
> Hi,
>
> First, I appreciate your tone of constructive criticism — there has been way
> too much negative criticism on this topic. However, as another user who has
> his own experiences with pgAdmin4, I feel compelled to reply regarding some
> of your complaints — please see my comments below.

+1. Thanks Jack.

>
> I don’t doubt that you are experiencing this, but I do not experience
> increased sluggishness vs. pgAdmin III at all. The difference, I believe, is
> that I am running OS X on a Mac, instead of Windows 10. So this may be a
> program tuning problem regarding Windows 10, or it may just be that the
> browser rendering in Windows 10 is poor for what the developers are doing.
> Either way, this can be a clue as to how to address this issue.
>
> Question for the Developers: Are you all using Windows as your main testing
> platform, or are most of you users of other platforms? If you’re not using
> Windows, then that may account for the poor performance in Windows — since
> most “business” work is done on Windows, it behooves you to focus on Windows
> performance primarily (I say this even though I was thrilled to get off
> Windows for so many reasons).

All of the EDB team use either Mac or Linux (on a VM on Mac)
primarily. I believe the Pivotal team are all Mac users as well.

That said, I have been doing a fair bit of testing on Windows over the
last week - but have failed to reproduce the extremes of slowness I've
seen reported. For example, one user said he was using a modern quad
core machine with 16GB RAM and SSD disk, and saw 1 minute+ startup
times. I could only get close to this on a 10 year old Dell Optiplex
740 with a dual core AMD CPU, 6GB RAM and a spinning disk where I was
seeing about 50 seconds to startup.

On a 4GB, 2 core VM running on my Mac, I see ~20 seconds following a
couple of tweaks I committed over the weekend, and oddly on a quad
core i7 with 16GB and SSD, I'm seeing much the same startup time - so,
not as fast as it ideally should be, but also nothing like as slow as
some have reported.

My testing so far is indicating that the slow part is QtWebEngine, the
embedded Qt browser that's used in the runtime. Google shows that
other users have also found this to be slow on Windows. Unfortunately
I've yet to find a better option to replace it. This has been bourn
out by other users who have reported much improved performance by
running the server from the command line and connecting with a
browser.

That said, we're continuing to look at how performance there might be
improved, as well as in other areas; for example, we're working on
eliminating JS/CSS templates in the backend in favour of static files
(which Ashesh has almost finished). That will minimise the amount of
backend processing done when loading code. We'll also be webpacking
the JS/CSS code to massively reduce the number of round trips the
client makes to the server to load everything. This should also
eliminate the "first-click delay" seen when opening some treeview
nodes for the first time that occurs when it on-demand loads the
required code.

So I guess my main question here is; what is different about the OPs
"*very* powerful and modern x86 workstation" that makes pgAdmin run so
slowly on it? My first guess is anti-virus software. My machines all
use Windows Defender, but perhaps something else is slowing down
things (pgAdmin does have a lot of files to access). Perhaps adding
the pgAdmin installation directory to the AV package's exclude list
would help.

> I have this same problem with pgAdmin III. 4 would not connect with my local
> server (where I have the empty password), so I cannot say if 4 has this
> issue.

I do not. pgAdmin 3 and 4 both remember my passwords just fine.

> My experience with 4 is much better than 3 here. 3 did not ever remember any
> context, so much so that if I had created a new server connection and got a
> timeout error and the program died when I tried to reopen the connection,
> the server would be lost and I would have to re-enter it’s properties
> (unless I first closed out of 3). 4 seems to at least not have that problem.
>
>
> While the fault for this probably lies in the Windows rendering engine, that
> does impact the choice of a browser-based implementation. Can you try this
> on a different browser, such as Firefox or Chrome (I don’t know if that’s
> even possible)?
>
> On Jun 11, 2017, at 4:11 AM, <grekloedlc(at)tutanota(dot)com>
> <grekloedlc(at)tutanota(dot)com> wrote:
>
>
> Dear pgAdmin developers,
>
>
> When I first heard that you were doing a total rewrite of pgAdmin III, I was
> extremely excited, because I had long been frustrated with the various
> annoyances and bugs in that program, which I was (and am still, actually)
> using daily, primarily how it always felt like a chore to start up and get
> ready unless the server was physically located near me.
>
>
> So when you eventually released the first public version of pgAdmin
> generation 4, I very eagerly downloaded it and tried it out. Unfortunately,
> to claim that I was "disappointed" isn't enough; I was frankly *appalled*. I
> will list the reasons in a moment, but I'd first like to point out that
> since then, I have repeatedly tried new versions of it, hoping to see
> improvements, but unfortunately finding all the same issues still present
> (as of v1.5, 2017-06-11, at the time of typing).
>
>
> Here are the critical issues:
>
>
> 1. Extreme sluggishness. Both the GUI itself, and the fetching of data from
> the server, is so slow and flimsy as to drive me insane even efter using it
> for less than a minute. I seriously get so angry that I kill the window in
> disgust. I'm not trying to be insulting or overly dramatic; this is just a
> fact. The software makes me angry due to how slow and unreliable it seems,
> and how prone it is to freeze (although it recovers after a while). This is
> on a *very* powerful and modern x86 workstation running a very "clean"
> (relative) Windows 10. All other programs are responsive and fast, except
> for pgAdmin 4. I really feel handicapped using it, in a way that's not at
> all the case with the old pgAdmin III, although even that one has mysterious
> fetch-delays that don't seem to correspond with the amount of data pulled
> through the network (SSH tunnel)…
>
>
> 2. It doesn't remember the empty password. It just keeps on asking, again
> and again, for the nonexistent password, even though I've checked the box to
> "remember" it a million times. This is infuriating to say the least.
>
>
> 3. Even worse so than the old program, pgAdmin 4 also doesn't seem to
> remember the "last state" at all, forcing me to slowly progress through the
> tree hierarchy each time I start it, waiting seconds each time I click
> anything. This makes me just let out a big sigh each time I have to manage
> my databases in any way, including making simple queries in a graphical
> environment. What should be instant becomes a huge chore. I cannot believe
> that it doesn't remember the "state" of the collapsed objects until the next
> time.
>
>
> In order to say something positive, I do appreciate the cross-platformness
> and apparently the ability for it to run in a browser, hosted on a server.
> (Although I personally don't trust it or any other software to do that
> safely.) Sadly, this has the serious downside of extremely poor performance,
> at least on Windows, to the point of making it practically impossible to
> use.
>
>
> I don't wanna sound as if I'm just telling you what a terrible job you've
> done. I realize that in spite of these serious flaws, a lot of work must
> have been plowed into this project, and it's unlikely that my complaints
> will really be taken to heart by the people who worked on it for so long,
> and for free, only to then get "insulted". I feel genuinely sorry and
> frustrated about the whole situation, and I'm now seriously wondering what
> to do with my "computer life" as it is heavily dependent on PostgreSQL as
> the basis. pgAdmin III is aging and pgAdmin 4 doesn't seem to be going
> anywhere, or changing in any major ways from its current state.
>
>
> What are the odds that you'll forget about pgAdmin 4 and instead go straight
> for a "pgAdmin V", taking everything you've learned but improving on it
> heavily? By the way, it is extremely common for developers to first do
> something great, then try to improve it, but failing entirely, instead
> producing a monster. For example: Winamp. There are many more cases, and it
> seems to happen again and again. It even happened to me! I was super proud
> of a product that was, to me, "vastly superior" to the old one, but the
> users absolutely hated it, and eventually, I had to realize that while
> technically better in some aspects, I had just done things "differently for
> the sake of doing them differently". I hope you'll understand me and that I
> really just want a great pgAdmin tool -- not to be mean.
>
>
> If you have anything promising to tell me in regards to any of this, I'd
> like to hear it. In the past, I've looked through the miserable
> "alternatives", so it's probably pointless to tell me about any of those,
> but if there is some sort of alternative that you know of, which is heavily
> polished and maintained and trusted and free of charge, it would definitely
> be interesting to me. However, I very much doubt that anything like that
> exists, and I doubt that this is the best place to ask for that. In fact,
> it's probably considered rude...
>
>
> // A long-time pgAdmin user who'd hate to see this crucial tool go the same
> way as so many other now-dead programs.
>
>

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise PostgreSQL Company