Skip banner and navigation tools.

 |  site map

swifttools.ukssdc.xrt_prods v1.10 Release Notes

Version 1.10 of the xrt_prods module was released as part of swifttools v3.0 on XXXXXX

This release features a number of important changes mainly relating to the way in which data products are returned and formatted. We have made significant efforts to support backwards compatibility, so all of your existing scripts should continue to work in the present release, provided you add one line at the top of your code. However, this will not be the case indefinitely: the affected old behaviour is deprecated and will be removed in a future release.

The simplest way to force the old behaviour is to set the XRTProductRequest.useDeprecate variable to True, which you can do straight after importing. So if you have followed the previous documentation for styling you will change to:

In [1]: from swifttools.xrt_prods import XRTProductRequest
In [2]: XRTProductRequest.useDeprecate = True

More details on managing deprecations appear below.

As well as changing certain behaviour, we have also introduced a number of new features; these are briefly described below but discussed in more detail in the main API documentation.

We have provided a Jupyter notebook with a hands-on whistle-stop tour of the changes for those who prefer to learn by use, rather than reading documenation.

Contents / change list

New functionality

xrt_prods v1.10 includes a few new features which I hope will be of use to you. I like them already. These are introduced in the next two sections, but documented more completely in the main documentation (as linked to below).

Spectral fitting controls

In August 2022 the XRT product generator was modified to give greater control over the automated fits to the generated spectra. v1.10 brings support for these changes to the xrt_prods API as well. These changes allow you to completely skip the spectral fitting phase, to control whether an absorber fixed at the Galactic value is included, to set the significance of the parameter errors, and to choose from among 3 possible emission models. The new parameters for a spectrum are:

These are discussed fully on the How to request products page

Note Use of these features requires using the new data structures introduced in v1.10 of xrt_prods. If you make use of these new features while also setting useDeprecate=True then, well I don't actually know what will happen, but it probably won't be pretty. Don't try it.

Saving spectra

We have added a new function saveSpectralData(), which allows you to download the data related to your created spectra. You can select to save only the png images of the spectra, or the actual data, as a tar.gz archive which you can also select to extract. This is more fully documented on the How to retrieve the products page.

This is in addition to the downloadProducts() function which is unchanged.

Light curve manipulation

Two new features have been included which let you do more with your light curves. The first is the rebinLightCurve() function, which lets you rebin the lightcurve without having to rerun the whole product build, it just redoes the binning phase of the light curve. The second is a function called mergeLightCurveBins(). This allows you to, er, merge bins in a light curve. For example, you may have created a light curve with a bin per observation, but your source had an extended period in which this yielded no detection, and you want to merge the resultant upper limits (either to get a deeper limit, or see if this merging yields a detection).

Both of these are shared functions within the swifttools.ukssdc module and so iare documented on the common functions page rather than here, but it will work with light curves created and retrieved using the xrt_prods module.

Note mergeLightCurveBins() has certain expectation of the light curve you give it, and it will not work on a light curve retrieved with deprecation enabled. So if you did give the useDeprecate=True argument then the light curve you obtained will not be properly handled by this new function. If you want to use the v1.10 functionality, you need to use v1.10, rather than forcing older behaviour.

Managing the deprecated features

As just noted above, the quick-and-easy solution to not caring about the changes is to tell xrt_prods to use the deprecated features. You can do this in a few ways (the code snippets below assume you've imported the XRTProductRequest object).

In [1]: XRTProductRequest.useDeprecate = True
In [2]: myReq = XRTProductRequest('YOUR_EMAIL_HERE', useDeprecate=True)
In [3]: myReq.deprecate=True

Obviously, running all 3 above would be a bit pointless since they all do the same thing, I'm just showing the options. In my opinion the top option is the best one, because it is a simple one-off entry at the top of your script/notebook/whatever and affects everything in the script. The second and third options obviously affect only a single request, and demonstrate changing the setting during the constructor or after the case. Of course, it is possible to set the variable to False instead of True.

While you are using the deprecated behaviour, you will get warnings appear the first time a script calls one of the deprecated functions. After about 11 seconds you may find this annoying, so you can turn them off by setting the (new) showDepWarnings variable in an XRTProductRequest object to False. You can do this when creating the object or subsequently, as in the two examples calls below:

In [4]: myReq = XRTProductRequest('YOUR_EMAIL_HERE', showDepWarnings=False)
In [5]: myReq.showDepWarnings=False

(Of course, calling both of these would be overkill).

Frankly, however, if you are going to modify your code to stop all those pesky warnings from appearing, it would be better to modify it to stop using deprecated behaviour. The needed modifications are generally pretty small, so let's take a look at them.

Deprecated functionality

The changes in functionality are pretty minor, and almost entirely relate to the way in which you retrieve products that you have requested. I've added a quick note about why this change exists at the end of this section, for the curious, but let's dive straight into exploring the changes. We'll start with a really trivial one and then move on to the slightly more impactful things.

Importing xrt_prods

The xrt_prods module has been moved inside the new swifttools.ukssdc module, and so "should" be imported with

In [6]: from swifttools.ukssdc.xrt_prods import XRTProductRequest

Or if you prefer to keep your namespaces clean:

In [7]: import swifttools.ukssdc.xrt_prods as ux

(You don't have to use the as ux but it's a bit shorter, and also follows a convention for all the swifttools.ukssdc imports).

You do not have to do this. I have created a wrapper file, so the module swifttools.xrt_prods still appears to exist, and I have no current plans ever to change this, but I'd still advise it partly in case I ever change my mind, and also because it makes all the new imports look more homogenous and pretty.

Retrieving data products

The XRTProductRequest class includes functions for retrieving the data products you built, but the behaviour of these differed from product to produce. All of the functions would return the retrieved data, but retrieveLightCurve() and retrieveSpectrum() would also store the retrieved data inside a class variable (myReq.lcData and myReq.specData).

From v1.10, this behaviour has been homogenised and slightly altered. Now all retrieve...() functions:

You can still ask the function to return the data, by passing the (new) argument returnData=True, e.g.

In [8]: lc = myReq.retrieveStandardPos(returnData=True)

and/or you can access the returned data via the class variables (most of which are new):

Note that these are class variables, i.e. they belong to your XRTProductRequest object, so if you called this myReq as in all my demos, you would prepend myReq. to the above variable names to access them.

Retrieving light curves

There is an extra small change to the retrieveLightCurve() function. The incbad and nosys arguments can now be "yes", "no" or "both" (or still True and False as before); i.e. instead of having to choose between only getting data with/out unreliable bins and with/out WT systematic errors, you can now get both. If you're wondering how you know what is what in your dataset, then read on…

Data structures

The format of light curve and spectral data that you retrieve via the retrieveLightCurve() and retrieveSpectralFits() functions has been revised somewhat. The object returned in each case is still a dict but with some changes to the structure compared to v1.9. These new structures are common across the entire swifttools.ukssdc module, and are described in detail on a dedicated page; the key changes compared to v1.9 are:

The column name changes are the only details I will explain here, for all else, please read the description of the new structures.

The changes to column names are all for purposes of homogenisation: the capitalisation of column names was previously all over the place, and the way in which error columns were labelled were likewise different in different places. From v1.10, all column labels are in CamelCase, and error columns end with "Pos" or "Neg". So, for example:

Also, for light curves consisting of upper limits, the "Rate" column has been renamed "UpperLimit" to help avoid confusion.

Parameter name changes

When you request a light curve, the timeType parameter has been replaced with timeFormat. timeType still works and is rewritten to timeFormat if you use it.

Why the changes?

For those of you cursing me, wondering why I have made you change your scripts, I can only (and sincerely) apologise: I hate it when people break backwards compatibility, even if they do it via deprecation, and having now done so myself I am the subject of my own contempt. I have persuaded myself that it was necessary (obviously, or I wouldn't have done it) and for anyone who cares, or can be persuaded to forgive me, I will briefly explain.

There were really three things at stake here.

  1. The creation of the swifttools.ukssdc module.
  2. The realisation that the existing behaviour was not very homogeneous.
  3. The need to support new functionality.

These are somewhat interconnected so I will just explain (briefly) prosaically.

When creating the new ukssdc module, with features to access data (including light curves and spectra) for GRBs and the SXPS catalogues, I intended to make the behaviour identical to that of the xrt_prods module which I hope you will agree was a laudable goal. It quickly became obvious that the xrt_prods behaviour was, frankly, rubbish. Column names had no consistency in their formatting (for historical reasons, specifically: "2007 me was an idiot."), some functions saved data to class variables and others didn't, and so on. I also realised that there were things I wanted to be able to do in this new module which the xrt_prods module didn't support -- such as the incbad='both' option for getting light curves, and for the reasons above it made sense to port these options to the xrt_prods module as well. For spectra, the SXPS catalogues contain two different spectral model fits, so the structure in which these fits are returned needed adjusting. I did consider not porting this new structure to xrt_prods, but instead took the opportunity to implement a long-promised feature for the product generator: the ability to fit more than just power-laws. Having done this, it was obviously essential that xrt_prods uses the new structure.

Having persuaded myself that these changes were justified, I have made considerable effort not to immediately break anyone's scripts, by adding a ridiculous amount of code and the front and back ends to ensure that users still on v1.9 or earlier get back from the server the data format they were expecting, and users on v1.10 can request the "deprecated mode", in which case the new-format data come back from the server, and then are internally "rewritten" to the old format.