New Light Curve Generator(LCGv2): Updates, Questions, & Suggestions
Announcement: New Forums
We are excited to announce the launch of our new forums! You can access it forums.aavso.org. For questions, please see our blog post. The forums at aavso.org/forum have become read-only.
American Association of Variable Star Observers (AAVSO)
Mon, 01/09/2017 - 13:03
This topic will report updates/improvements to the LCG. Also, you can post questions you may have, and offer any suggestions that you believe would improve it for your use.
The short answer is no. LCGv2 does not have this capability yet. However, I am in communication with Francis about this and hopefully he will be able to implement something along those lines.
Hi Tim (and anyone else interested in this capability),
Thanks to LCGv2's developer, Francis Hemsher, we now have available a new feature which will allow you to request a specific lightcurve via query string. In other words, by typing the following into the URL address box on your browser:
... you can get a fully interactive plot of R Sct!
There are also options to request a simple (non-interactive) plot, to specify what bands you would like to display, and to use JD or calendar date along the X-axis of the plot.
Also, you can see both (interactive and static) versions of the URL needed to duplicate a plot you create in the "Preferences" window as shown here:
Just select the link which you can then copy & paste into a webpage, send by email to a friend, or build into your own software programs.
First off, I really like the new LCG. The interactive capabilities are wonderful and very useful. I'm still using the old LCG for some things, however, because the API for the new one still isn't quite as flexible. One thing that would really help would be "Plot previous days" parameter. It's exposed in the interactive form for generating a new light curve, so it seems like it shouldn't be too hard. I'd like to be able to say "show the last 200 days," rather than having to compute the current JD/date and current-200 to pass in the URL.
We have included a URL link address for the specific light curve you have created in the LCG. This is shown under 'Preferences'. Two URLs are shown: Static and Interactive.
The Static URL provides just the image of the light curve. The Interactive URL provides the image of the light curve with all typical LCG user selections.
Note that the Static URL can be used anytime you want to show an image of your light curve in any application, journals, reports, emails, etc. It automatically adjusts its size calling environment.
When looking at a plot, if you turn off 'all contributors' and only select one contributor with error bars enabled it does not remove the error bars for the un-plotted data - it removes the points but leaves the bars.
Try this: Plot LC for star V1454 Cyg for (say) the last 5 days (as of writing this).
Near JD 2458013 you will see a high cadence time series by observer ATE, but this observer is not listed in the box under the plot.
EDIT: After some additional experimentation, I begin to suspect that there is always one observer missing from the box no matter what you input, perhaps the one who's first in alphabetical oder.
If you click on a data point to display the details in the "Observation data" box, and then click on a "fainter than x" data point, not all items in the "Observation data" box refresh to show the data from the new point, instead you get a mix of entries from the currently selected (fainter-than) point and the previously selected point. I just tested this also on a LC of V1454 Cyg.
One of the annoying "issues" is it doesn't seem to "save" my defaults that I like, when in the select star box. For example, I prefer Calendar day, not Julian day, and I prefer specific bands (V, Vis, CV) rather than all bands, abd a specific number of days backwards in time, but I always have to take the extra steps to reset these, because the defaults provided when I open the box, are the opposite!
I think there is an error in showing light curves using calendar date. I've attached an example LC of N2018 Per (V392 Per). It looks like one day shift on calendar scale, for instance instead of "2018/05/05" it should be "2018/05/06" date, and so on ...
I have date format yyyy-mm-dd in my Win10 settings.
Thank you for your post and the screenshot showing the problem you observed with the LCG.
Unfortunately, I am having trouble repeating the problem you reported (see attached screenshot). If you are able to repeat this problem again would you please go through step by step what you did?
For example, did you start out by requesting calendar date from the "Plot another curve" window or did you request the data in JD then switch to Calendar date using the "Preferences"? Did you zoom in to rescale the plot?
I generate LC of V392 Per from main AAVSO page using option "Plot a light curve" and the next step I change parametres (Julian Day --> Calendar Date, From Date ..., To Date ...) of LC in LCGv2 window. But probably this is not the problem.
When I take 12 days of LC of V392 Per I will get calendar date scaled to 22 UT on each day (this is my current computer time at the moment of the first static LC generation, but it is UT+2h), for instance "2018/05/03 22:00" - see below a static link to LC and attached screen shot 12daysLC.png .
When I take 13 days of LC of V392 Per I will get calendar date scaled without time on each day, for instance "2018/05/03" (time=22UT in date "2018/05/03" stays hidden???) - see below a static link to LC and attached screen shot 13daysLC.png.
To show the error bars using LCGv2 just click "Preferences" then under "Plot Preferences" click "Show Error Bars". If the drop-down box reads "> 0.0" you will see all error bars, but you can limit it to showing only the bigger ones if you want. Because the symbols are so large it can be hard to see the error bars sometimes.
everything becomes sooooo slooow with large data sets in LCGv2
Is there a simple way to speed up the user interface when a large data set is loaded. Even a simple click anywhere takes forever to process, as if it is thinking "I've downloaded a huge amount of data, let me think about that ... Oh! theres a user click!" Is everything cached and reloaded over and over?
Update: Yes, this is a problem with the browser. Changing from Firefox to Chrome vastly improves performance. My firefox is in such a bad state that it randomly forces my machine to shut down (I'll blame firefox at this point in my investigations anyway).
Another update: I had a fan problem and my computer died (RIP). So I take back what I implied about Firefox. I guess things are slow because everything has to be replotted, and speeding it up would require some tweaks in the plotting method or a faster computer.
I have 2 photos of an area next to CV Gru, took 2 months appart. In one of them appears a star (mag 8.5) that does not appear in the other. Can anyone examine these 2 photos? For as I saw this star does not appear in the AAVSO chart with mag. up to 11.0 or 12.0!
Hi Carlos,
This is the Software development forum and this thread is about the LCGv2 not about observations, so post in the General dicussion forum or in the Campaigns and Observation reports forum.
Your images have not been uploaded. Upload them as jpgs if possible. And determine coordinates for the object would always a better way to report en event like this so everybody can check.
Thanks for pointing out this problem, and your suggested code changes...I'm very impressed :)
Rather than scale the error bar lines during Zoom as you suggested, I recomputed the line lengths. Sara Beck is currently testing the change and will include them in the live site if they do the job.
Thanks much,
Francis
UPDATE: This change is now on the live site...give it a try!
When the display is zoomed more than once, and then the back button clicked, the body of the display is not zoomed to the same amout as the panning and display box at the top thinks. I also managed to get the JD display direction reversed whilst playing: labeled JD going down as you move to right and panning buttons working in reverse - but I haven't figured out how to reliably reproduce that problem.
It would be nice to be able to use the mean curve button when the display is zoomed.
Frustrating it still cannot handle large datasets :(
I frequently need to check the long term history of stars, and inevitably get the "Greater than 20,000 observations, some features disabled" error. The app then seriously slows down, and I can't see the observers list. I think many others use it too, where a larger number of observations will result, so this remains a very irritating limitation with this app.
It seems to me the program should have been written in a way to be scalable with the dataset size, since will become more of an issue as time goes by!
Could someone please fix the sizing issue? I like many of the LCGv2 features, but still must use LCGv1 to get around this problem, which shouldn't be necessary :(
The Observation Data box, displayed when an observation in the plot is selected, displays the details of that observation, including at the bottom, a "comments" box. That box is (almost) always empty, even if the details of that observation shown in the WebObs Search page contains a Note. The Note added by the observer to this observation is not shown elsewhere in the Observaton Data box in LCGv2.
I am able to add comments to the LCGv2 Comments box for any observation, include observations made by other observers. My added comments "stick" when I close and then reopen that observation whether it is mine or someone else's. Perhaps that was the LCGv2's author's intent.
I think it would be helpful if the LCGv2 Comments box contained the Note attached to that observation by the observer. I also suggest that only the observer should be able to edit that box.
Thank you for pointing out this bug. I have passed it on to the programmer with the suggestion that the comments box be read-only (along with all the other data displayed in the "Observation Data" window). Fortunately, there is no mechanism for LCGv2 to write anything back to our database so the actual data are really safe.
LCGv2 has lots of "nifty features", "bells and whistles", compared to LCGv1, but the slowdown and loss of functions with larger datasets is very annoying, limits its usefulness, especially for research and publication. If the developer cannot fix this, due to limitations of the platform, I think it needs to be re-written in a more robust, performance oriented and lower level language, like C++ or such.
It appears that LCGv2 has recently developed a couple of problems. First off, clearing the "highlight observations" checkbox next to an observer's code in the contributors table also clears the "show observations" checkbox between the obscode and name. The second issue is that the direct URL to plot data that was given in Sara's post back on 6/27/17 no longer works. I use that quite a lot as a custom Chrome search engine entry but now the web page only renders the top banner and an eternally turning spinner. The developer console reports "Uncaught TypeError: Cannot read property 'append' of undefined" at buildPlot.js:175.
I'm using Chrome 74.0.3729.169 on CentOS7, but see the same two problems using Firefox 60.6
Hi Tim,
The short answer is no. LCGv2 does not have this capability yet. However, I am in communication with Francis about this and hopefully he will be able to implement something along those lines.
Thanks, Sara
Hi Tim (and anyone else interested in this capability),
Thanks to LCGv2's developer, Francis Hemsher, we now have available a new feature which will allow you to request a specific lightcurve via query string. In other words, by typing the following into the URL address box on your browser:
https://www.aavso.org/LCGv2/index.htm?DateFormat=Julian&RequestedBands=&view=api.delim&ident=R Sct&fromjd=2457201.875086794&tojd=2457931.875086794&delimiter=@@@
... you can get a fully interactive plot of R Sct!
There are also options to request a simple (non-interactive) plot, to specify what bands you would like to display, and to use JD or calendar date along the X-axis of the plot.
Also, you can see both (interactive and static) versions of the URL needed to duplicate a plot you create in the "Preferences" window as shown here:
Just select the link which you can then copy & paste into a webpage, send by email to a friend, or build into your own software programs.
Your feedback is most welcome!
-Sara
VSX now also links out to the new LCG from the detail sheet of an object using this URL.
Patrick
First off, I really like the new LCG. The interactive capabilities are wonderful and very useful. I'm still using the old LCG for some things, however, because the API for the new one still isn't quite as flexible. One thing that would really help would be "Plot previous days" parameter. It's exposed in the interactive form for generating a new light curve, so it seems like it shouldn't be too hard. I'd like to be able to say "show the last 200 days," rather than having to compute the current JD/date and current-200 to pass in the URL.
Shawn (DKS)
We have included a URL link address for the specific light curve you have created in the LCG. This is shown under 'Preferences'. Two URLs are shown: Static and Interactive.
The Static URL provides just the image of the light curve. The Interactive URL provides the image of the light curve with all typical LCG user selections.
Note that the Static URL can be used anytime you want to show an image of your light curve in any application, journals, reports, emails, etc. It automatically adjusts its size calling environment.
I just noticed a nit with error bars.
When looking at a plot, if you turn off 'all contributors' and only select one contributor with error bars enabled it does not remove the error bars for the un-plotted data - it removes the points but leaves the bars.
Try this: Plot LC for star V1454 Cyg for (say) the last 5 days (as of writing this).
Near JD 2458013 you will see a high cadence time series by observer ATE, but this observer is not listed in the box under the plot.
EDIT: After some additional experimentation, I begin to suspect that there is always one observer missing from the box no matter what you input, perhaps the one who's first in alphabetical oder.
Cheers
HB
Looks like another issue:
If you click on a data point to display the details in the "Observation data" box, and then click on a "fainter than x" data point, not all items in the "Observation data" box refresh to show the data from the new point, instead you get a mix of entries from the currently selected (fainter-than) point and the previously selected point. I just tested this also on a LC of V1454 Cyg.
CS HB
One of the annoying "issues" is it doesn't seem to "save" my defaults that I like, when in the select star box. For example, I prefer Calendar day, not Julian day, and I prefer specific bands (V, Vis, CV) rather than all bands, abd a specific number of days backwards in time, but I always have to take the extra steps to reset these, because the defaults provided when I open the box, are the opposite!
Mike
When Preferences>Error Bars is checked the error bars appear in the initial plot, but they do not appear inside a zoomed in Box.
I think there is an error in showing light curves using calendar date. I've attached an example LC of N2018 Per (V392 Per). It looks like one day shift on calendar scale, for instance instead of "2018/05/05" it should be "2018/05/06" date, and so on ...
I have date format yyyy-mm-dd in my Win10 settings.
CS
Richard (BRIA)
Hi Richard,
Thank you for your post and the screenshot showing the problem you observed with the LCG.
Unfortunately, I am having trouble repeating the problem you reported (see attached screenshot). If you are able to repeat this problem again would you please go through step by step what you did?
For example, did you start out by requesting calendar date from the "Plot another curve" window or did you request the data in JD then switch to Calendar date using the "Preferences"? Did you zoom in to rescale the plot?
Many thanks,
Sara
Hi Sara,
thank you for your tests V392 Per LC generation.
I generate LC of V392 Per from main AAVSO page using option "Plot a light curve" and the next step I change parametres (Julian Day --> Calendar Date, From Date ..., To Date ...) of LC in LCGv2 window. But probably this is not the problem.
When I take 12 days of LC of V392 Per I will get calendar date scaled to 22 UT on each day (this is my current computer time at the moment of the first static LC generation, but it is UT+2h), for instance "2018/05/03 22:00" - see below a static link to LC and attached screen shot 12daysLC.png .
Static link 12 days LC
When I take 13 days of LC of V392 Per I will get calendar date scaled without time on each day, for instance "2018/05/03" (time=22UT in date "2018/05/03" stays hidden???) - see below a static link to LC and attached screen shot 13daysLC.png.
Static link 13 days LC
CS,
Richard
Hi Richard,
I think that the problem has been fixed. Please check and let me know what you think.
Many thanks to you for finding this and to Francis (LCGv2's developer) for fixing it!
-Sara
LCG1 is broken.
How do I get error bars on LCG2?
Ray
Hi Ray,
To show the error bars using LCGv2 just click "Preferences" then under "Plot Preferences" click "Show Error Bars". If the drop-down box reads "> 0.0" you will see all error bars, but you can limit it to showing only the bigger ones if you want. Because the symbols are so large it can be hard to see the error bars sometimes.
I hope that helps.
Sara
The size of error-bars is not adjusted when the display is zoomed.
William
Is there a simple way to speed up the user interface when a large data set is loaded. Even a simple click anywhere takes forever to process, as if it is thinking "I've downloaded a huge amount of data, let me think about that ... Oh! theres a user click!" Is everything cached and reloaded over and over?
Update: Yes, this is a problem with the browser. Changing from Firefox to Chrome vastly improves performance. My firefox is in such a bad state that it randomly forces my machine to shut down (I'll blame firefox at this point in my investigations anyway).
Another update: I had a fan problem and my computer died (RIP). So I take back what I implied about Firefox. I guess things are slow because everything has to be replotted, and speeding it up would require some tweaks in the plotting method or a faster computer.
William
I have 2 photos of an area next to CV Gru, took 2 months appart. In one of them appears a star (mag 8.5) that does not appear in the other. Can anyone examine these 2 photos? For as I saw this star does not appear in the AAVSO chart with mag. up to 11.0 or 12.0!
Thanks, Carlos
Hi Carlos,
This is the Software development forum and this thread is about the LCGv2 not about observations, so post in the General dicussion forum or in the Campaigns and Observation reports forum.
Your images have not been uploaded. Upload them as jpgs if possible. And determine coordinates for the object would always a better way to report en event like this so everybody can check.
Cheers,
Sebastian
Thanks, Sebastian!
This can be fixed by setting (in 10_errorBar.js)
.attr("y1", function(d)
{
return -d.uncert // replacing: return -d.uncert*magXFactor
}
)
.attr("y2", function(d)
{
return +d.uncert // replacing: return +d.uncert*magXFactor
}
)
and replacing: (in 05_boxPlot.js, 07_boxSlide.js, and 10_errorBar.js)
ErrorBars.attr("transform", function (d)
{
return "translate("+XscaleJulian(d.JD)+" "+YL(d.mag)+")"
}
)
with:
var scaleY = YL.domain()[0] - YL.domain()[1]
ErrorBars.attr("transform", function(d)
{
return "scale(1, "+scaleY+")"+"translate("+XscaleJulian(d.JD)+" "+YL(d.mag)/scaleY+")"
}
)
This does not fix error bar scaling when you change magnitude range in Light Curve Preferences.
Also, when zoomed in you can see that the position of points is offset slightly in magnitude, and 3 decimal digits of jd precision would be useful.
Willam
Hi Willam,
Thanks for pointing out this problem, and your suggested code changes...I'm very impressed :)
Rather than scale the error bar lines during Zoom as you suggested, I recomputed the line lengths. Sara Beck is currently testing the change and will include them in the live site if they do the job.
Thanks much,
Francis
UPDATE: This change is now on the live site...give it a try!
Your fix is much better than mine, thankyou.
William
When the display is zoomed more than once, and then the back button clicked, the body of the display is not zoomed to the same amout as the panning and display box at the top thinks. I also managed to get the JD display direction reversed whilst playing: labeled JD going down as you move to right and panning buttons working in reverse - but I haven't figured out how to reliably reproduce that problem.
It would be nice to be able to use the mean curve button when the display is zoomed.
William
I frequently need to check the long term history of stars, and inevitably get the "Greater than 20,000 observations, some features disabled" error. The app then seriously slows down, and I can't see the observers list. I think many others use it too, where a larger number of observations will result, so this remains a very irritating limitation with this app.
It seems to me the program should have been written in a way to be scalable with the dataset size, since will become more of an issue as time goes by!
Could someone please fix the sizing issue? I like many of the LCGv2 features, but still must use LCGv1 to get around this problem, which shouldn't be necessary :(
Mahalo,
Mike LMK
The Observation Data box, displayed when an observation in the plot is selected, displays the details of that observation, including at the bottom, a "comments" box. That box is (almost) always empty, even if the details of that observation shown in the WebObs Search page contains a Note. The Note added by the observer to this observation is not shown elsewhere in the Observaton Data box in LCGv2.
I am able to add comments to the LCGv2 Comments box for any observation, include observations made by other observers. My added comments "stick" when I close and then reopen that observation whether it is mine or someone else's. Perhaps that was the LCGv2's author's intent.
I think it would be helpful if the LCGv2 Comments box contained the Note attached to that observation by the observer. I also suggest that only the observer should be able to edit that box.
Phil
Hello Phil,
Thank you for pointing out this bug. I have passed it on to the programmer with the suggestion that the comments box be read-only (along with all the other data displayed in the "Observation Data" window). Fortunately, there is no mechanism for LCGv2 to write anything back to our database so the actual data are really safe.
-Sara
Hi Sara,
I think it would be useful to add the observer's Notes from the original data report file to the Comments box, if this is possible.
Phil
LCGv2 has lots of "nifty features", "bells and whistles", compared to LCGv1, but the slowdown and loss of functions with larger datasets is very annoying, limits its usefulness, especially for research and publication. If the developer cannot fix this, due to limitations of the platform, I think it needs to be re-written in a more robust, performance oriented and lower level language, like C++ or such.
Mike
It appears that LCGv2 has recently developed a couple of problems. First off, clearing the "highlight observations" checkbox next to an observer's code in the contributors table also clears the "show observations" checkbox between the obscode and name. The second issue is that the direct URL to plot data that was given in Sara's post back on 6/27/17 no longer works. I use that quite a lot as a custom Chrome search engine entry but now the web page only renders the top banner and an eternally turning spinner. The developer console reports "Uncaught TypeError: Cannot read property 'append' of undefined" at buildPlot.js:175.
I'm using Chrome 74.0.3729.169 on CentOS7, but see the same two problems using Firefox 60.6
Shawn
Thank you for the bug reports Shawn. Fixes for both are in work.
-Sara