GitLab

Fixed various notices throughout LOVD

We used to do quick checks on $_AUTH['level'], but $_AUTH can also
 be false. This now triggers notices in newer PHP versions.
 Replacing all calls to fix these notices. Starting with the ajax
 scripts.

Improve `lovd_getVariantInfo()` and `lovd_fixHGVS()`

Removed some unneded code, resorted two tests, re-added one test.

Positions in the 3' UTR weren't handled well yet.

Added tests for lovd_fixHGVS() that were recently added for
 lovd_getVariantInfo().

The first is extracted and processed; then the type is overwritten
 by ";" and an ENOTSUPPORTED is added. Possible warnings that can
 occur from correct HGVS descriptions of combined variants are
 removed. This suffices for now.

These now throw an ENOTSUPPORTED as long as they don't match the
 regex; otherwise they already threw a WSUFFIXGIVEN. We currently
 don't prevent this as that's for a later step when we decide to
 properly support them. At least like this, we'll get positions and
 we can recognize them and therefore allow these variants to be
 entered in LOVD.

These are variants taken straight from VCF fields, like g.100A>AT
 that should be g.100_101insT, or even more complex variants like
 g.100ATGA>AA that should be g.101_102del.

Empty ALTs are supported, so g.100A>. becomes g.100del.

However, empty REFs are not supported as it's unclear where the
 insertion should take place. Either way, an empty ALT is not a
 valid VCF file.

... based on a start position (given in the format of
 lovd_getVariantInfo()) and a variant length (1 being the minimum).

These are constructed from VCF files and are translated into
 substitutions. The NN>N or N>NN types we already recognize. Now
 also recognize variants where either the REF or the ALT was left
 empty, like N>. or even .>N - both of which are actually bugs in
 the VCF file generator.

g.?_(?_?)insAAAA can still be fixed.

This is now a bit simplistic; the main regex is now case-insensitive,
 the prefix is forced to be lowercase, and the variant type is forced
 lowercase unless for substitutions, WT variants, and repeats -
 those are forced into uppercase. For bsrC an exception is made.

The suffix is left as is and is not checked.

Now separately checking full positions passed with refseqs.

It was running regexes over the whole suffix, which sometimes led to
infinite recursion. Now, the suffix is split and all parts are checked.
This fixes the infinite recursion and will also allow us to check
suffixes better.