`isolationWindow` fails with libxml2 2.14+

[rekado]

The XML package depends on a feature in libxml2 to transparently decompress gzipped files. As of libxml2 2.12 this feature is only available when configured with --with-legacy. As of libxml2 2.14 the feature no longer works even with the --with-legacy switch. Decompression via zlib is slated for removal in the next release.

isolationWindow in R/utils.R uses XML::xmlParse on potentially gzipped files. An example is the test in inst/unitTests/test_isolationWindow.R:

msdata::proteomics(full.names = TRUE,
                            pattern = "TMT_Erwinia_1uLSike_Top10HCD_isol2_45stepped_60min_01.mzML.gz")

The isolationWindow procedure fails in the call to XML::xmlParse when libxml2 2.14 is installed (or libxml2 2.12+ without the --with-legacy option) because the raw gzipped file is passed to the XML parser.

[lgatto]

Thank you @rekado for brining this up. I just see now that you have been tackling this in guix. I will read through the issue and try to see how to address this here, in mzR - apparently, it's not even here we could fix this. I'll try to report upstream.

Thank you again for this detective work!

[rekado]

I reported the bug here, because at first I thought it should check if the XML package reported the libxml2 zlib feature before passing a gzipped XML file to it. But it is a feature claim of the XML package, so it's reasonable to expect that they work around a missing feature in libxml2 instead of all users of the XML package.

[lgatto]

We mostly rely on proteowizard to parse the mass spectrometry data files (that use a specific XML schema). There's one exception, and that is isolationWindow(), that directly uses the XML package.

@rekado - do you know if the xml2 package addresses the changing/missing features?

[lgatto]

@sneumann @jorainer - we are also extracting isolationWindow parameters (target MZ, lower/upper offset), that have the following cvParams: MS:1000828 and MS:1000829. We could also compute these from the header data:

low = isolationWindowTargetMz - isolationWindowLowerMz
hight = isolationWindowUpperMz - isolationWindowTargetMz

Unless we have a simple fix to parse the mzML file:

• should we deprecate isolationWindow()?
• should we provide an alternative implementation and computer as suggested above?
• can we extract these cvParams from the mzML files using pwiz?

PS: @jorainer - note that we don't even have an isolationWindow() in Spectra.

[rekado]

@lgatto xml2 checks if the first character is <; if it isn't it will treat it as a connection and read it with https://github.com/r-lib/xml2/blob/main/R/xml_parse.R#L180.

If it's a gzipped connection it would be handled by R before passing it through to the XML document parser.