Primenet web design

[Madpoo]

Originally Posted by Uncwilly

I can obviously "fix" my macro to start 1 row further down. My concern is keeping things as "normal" as possible on PrimeNet (each cell behaves as a cell, headers are nearly always a single line, delimiters are used to indicate blank fields, etc.). What ever you do, don't create an <XML> just for me. My queries are just for tracking the 100M digit range.

That's weird that it was importing the CSV data funny.

I do have an XML version all set to go if you change you're mind. Looking at that page gave me some ideas about making the interface itself better (not the data or the format, but things like if you click the "output worktodo" it will auto-click the "exclude assigned" and "only TF" options. It does that automatically when it's submitted, but it's nicer to see that when you actually click the box in the first place so you know what to expect.

None of the table data (or CSV) layout would change. If I do release these tweaks, I'll be sure that the resulting output is unchanged from this point on. On the off chance it just *has* to change for some reason, I'll consult first and see if it'll cause problems.

As for the <br>, I can solve it by removing that and just setting a specific width on that column so it wraps on it's own. I'm not sure if, stylistically, that's a better or worse option... maybe 6 of one, 1/2 dozen of the other in the end.

[Uncwilly]

Originally Posted by Madpoo

As for the <br>, I can solve it by removing that and just setting a specific width on that column so it wraps on it's own. I'm not sure if, stylistically, that's a better or worse option... maybe 6 of one, 1/2 dozen of the other in the end.

Maybe we can hear some other voices.

If I saved the CSV and then imported the file I am sure that it would work fine. Why excel is too stupid to do that with a web query???? Again, someone called it the town idiot.

[Madpoo]

Originally Posted by Uncwilly

Maybe we can hear some other voices.

If I saved the CSV and then imported the file I am sure that it would work fine. Why excel is too stupid to do that with a web query???? Again, someone called it the town idiot.

Not sure why Excel has a hard time... the CSV output has two modes:
1) it displays the CSV in a code block on the regular page (technically a <pre> block...)
2) It can offer up a "report.txt" file to download, with just the CSV contents.

With option #1, if Excel tries to view that, there's still a lot of stuff on the page, but if Excel could look for that <pre> block and only parse that, seems like it shouldn't have a problem. Not sure.

With option #2, the "report.txt" is setup to deliver as text/plain, but as an attachment... if not for that, it would display the plain text in a browser and I presume Excel could also get that directly. As is, I don't think Excel would be able to do anything with the "download" option since it would probably (and rightly) not handle something like that?

I guess another option would be to change it so the text display shows just the text output and not all the other HTML. Plain, boring text on a page.

But that's where I kind of think XML is a better choice at grabbing data from the interwebs and pulling into something else. That or JSON.

I imagine that ideally, with all the data mersenne.org has on all kinds of progress on factoring, LL, DC, ECM, P-1, etc. etc. that an actual RESTful API would be a cool thing, with methods covering all the things people are interested in.

People like yourself are already grabbing stuff from these existing reports and while that can obviously work, it's super inefficient, on the server and the client ends. I like efficient client/server communication: it makes me happy.

[Madpoo]

Okay, this page now has an XML option and some other updates to the look up top.

The XML will output data like:
<factoring>
<data>
<exponent>1277</exponent>
<factorDepth>61</factorDepth>
<pm1Bound1>2000000000000</pm1Bound1>
<pm1Bound2>30000000000000</pm1Bound2>
</data>
</factoring>

Any existing things anyone had to pull reports from there can just use the same URL parameters, but add a URL parameter "xml=1" (and remove any "txt=1" it might have used if it was generating the CSV stuff previously).

The other things I changed were to add the table sorting theme so that arrows show up for sorting. Plus when you select certain options, other options that need to be checked or unchecked will automatically update (it's Javascript; sorry Retina, but that's the only way something like that will work)

[retina]

Originally Posted by Madpoo

The other things I changed were to add the table sorting theme so that arrows show up for sorting. Plus when you select certain options, other options that need to be checked or unchecked will automatically update (it's Javascript; sorry Retina, but that's the only way something like that will work)

I have no issue with this because I don't use the web browser to do things like sorting or other advanced magic. I've got a whole computer here that I can use to do these things if and when I desire it, the browser uses only a small portion of it. Hah, so there, I just said I don't care.

And yes (to answer the unasked question) I am aware that for some people the browser is the computer to them.

[kracker]

http://www.webpagetest.org/result/14...3fdea28b83c0e/

[Madpoo]

Originally Posted by kracker

http://www.webpagetest.org/result/14...3fdea28b83c0e/

And not just the home page load times, but the time it takes to load some of the top team/account reports, and the exponent reports... lots faster than it used to be on the old server.

I can't say that the new design of the website has made the speed any better or worse. Both of them used some javascript and images for navigation and layout so it's about the same page weight and # of resources. I like to tell myself that it *feels* better in the redesign but that's probably a subjective thing anyway.

I know some of y'all didn't care one way or the other how the site looks, but for the rest of the world who might be coming to the site for the first time, it should hopefully give a good first impression.

Especially when we find a new prime and the press release goes out, that page should look spiffy and load fast, and all the new visitors will say "Hey, that looks like a nice website, I think I'll sign up."

That's how it works, right?

[paulofaller]

As a user who complained about the website in February of this year and has only just found this thread after seeing the update go live...
Thank you for listening and the new design is significantly better, well at least from my perspective.

[Madpoo]

Originally Posted by paulofaller

As a user who complained about the website in February of this year and has only just found this thread after seeing the update go live...
Thank you for listening and the new design is significantly better, well at least from my perspective.

To save me the incredible hassle of looking back at your previous post </sarcasm> did you have any specific complaints or suggestions?

The stuff I've done was mostly my own bias with input and a basic thumbs up/down from George and James since they've been doing the care and feeding of the site, and I didn't want to muck up their process.

I've tried to take specific ideas or concerns into account when people mention it, and aside from a few things like adding the current UTC somewhere (like the footer) I think right now the redesign work is mostly done and I'm looking at any areas for new features.

One page in particular I'm looking at yesterday and today is the exponent report http://www.mersenne.org/report_exponent/

That one currently suffers from some overflow if there are some really long factors, so I've got a redesign mocked up to wrap long factors but make it readable still if there are more than 1 known factor. Perfect example is this bad boy:
http://www.mersenne.org/report_expon...o=1109&exp_hi=

That report is also very confusing (well, to me anyway) if you select to show the full history. It adds a history sub-table for each exponent and it basically makes the report so bulky and hard to ready, it's kind of unusable.

I'm exploring having the history be just a hover pop-up, or maybe only shows up if you're looking at just a single exponent.

It all comes down to use-cases... do people like to look at a report of a lot of exponents and have the full history of more than one exponent showing up at a time? Or if they *do* like those reports, do they need it in a table or are they getting just the "text" output anyway?

The end goal being to have a report that's easy on the eyes but still useful for whatever people do with 'em.

My other goal that's probably longer term is some kind of graph over time of an account or team's GHz-Hours throughput, like the aggregate one showing the TFlops of GIMPS as a whole.

Longer longer term is doing something with the benchmarks. Again, for me, it's kind of confusing to look at the numbers and figure out what to do with them. I mean, I know what they all are, but the data is presented in a way that makes it hard to pull meaning from it, and some of the data itself could use some work in general. Older Prime95 versions would report different features for the same CPU, which resulted in some strange branching or segmenting of results on the server and I'd like to flatten that out when possible. It could cut the # of distinct "cpu types" in half.

[James Heinrich]

Originally Posted by Madpoo

or maybe only shows up if you're looking at just a single exponent.

That would be my suggestion.

[S485122]

Originally Posted by Madpoo

...

That report is also very confusing (well, to me anyway) if you select to show the full history. It adds a history sub-table for each exponent and it basically makes the report so bulky and hard to ready, it's kind of unusable.

I'm exploring having the history be just a hover pop-up, or maybe only shows up if you're looking at just a single exponent.

...

For the bulk of exponents the history data only contains factoring and LL testing data. It is for the smaller exponents that one gets a lot of ECM data.

At the moment the "Exponent Status" report does not show all the content of the database about an exponent. To get more of the known data one must use the history, for instance the bad LL data is shown in the history only, meaning that data would not be shown for pre 2008 results : 21337531 Exponent Status. Notice that the display of numbers uses three different formats : commas as digit grouping symbol for the exponent, spaces for the P-1 bounds and finally no digit grouping for factors.

For exponents with pre 2008 bad results or for which an factor was found later one must combine two searches : 25273219 Exponent Status and 25273219 LL Test Results or 27804209 Exponent Status and 27804209 LL Test Results In the later report the date of the last test is missing even if it is present in History port of the former report. Another thing is that when showing LL results only the verified and unverified results are presented as a table, the bad results and "Test results where factor was found later" are shown as text.

If one could see all known data (factoring limit, P-1 bounds and all LL tests) about an exponent whatever its state (prime, unverified verified or factored) one could limit showing the history to single exponent reports. There is a catch : ECM data only seems to exist in the history (there used to be a report which showed a table of the different number of curves for ranges of bounds, perhaps it was in the pre V4 era...) The text only report could go on showing the history even for ranges.

But all this is nitpicking, the work you did is tremendous,

Jacob