Refactor internals for speed, fix bugs, bump to 1.8.7

DESCRIPTION

@@ -1,7 +1,7 @@
 Package: variantstring
 Type: Package
 Title: Functions for working with variant string format
-Version: 1.8.6
+Version: 1.8.7
 Authors@R: c(
     person("Bob", "Verity", email = "r.verity@imperial.ac.uk", role = c("aut", "cre"))
     )
@@ -12,10 +12,10 @@ License: MIT + file LICENSE
 Encoding: UTF-8
 LazyData: true
 RoxygenNote: 7.3.2
-Suggests: 
+Suggests:
+    readr,
     testthat (>= 3.0.0)
 Config/testthat/edition: 3
-Imports: 
+Imports:
     dplyr,
-    stringr,
-    tidyr
+    stringr

NAMESPACE

@@ -23,4 +23,3 @@ import(dplyr)
 importFrom(stats,na.omit)
 importFrom(stringr,str_c)
 importFrom(stringr,str_split)
-importFrom(tidyr,pivot_longer)

NEWS.md

@@ -0,0 +1,30 @@
+# variantstring 1.8.7
+
+## Bug fixes
+
+* `get_component_variants()` no longer errors on valid strings that combine a
+  phased heterozygous locus with a homozygous locus (e.g. `pfcrt:72_73:C|S_V`).
+  The homozygous allele is now correctly carried across all phased components.
+* `check_position_string()` now validates every distinct input. Previously, when
+  the input contained duplicates, an invalid string positioned after the
+  duplicates could be skipped and wrongly pass validation.
+* `check_variant_string()` and `check_position_string()` now report the reason
+  for failure in the error condition itself (via `conditionMessage()`), instead
+  of printing it to the console and raising an empty error. This makes the
+  reason available to calling code that catches the error.
+
+## Behaviour changes
+
+* `drop_read_counts()` and `position_from_variant_string()` now preserve the
+  format of the input rather than re-normalising it. In particular, position
+  ranges (e.g. `72-74`) are kept compact rather than expanded to `72_73_74`, and
+  the concise amino-acid notation (e.g. `CV`) is preserved rather than rewritten
+  as `C_V`. Genes are still returned in alphabetical order.
+
+## Internal
+
+* Substantial refactor of the internals for speed and to remove duplicated code,
+  with no change to the documented behaviour of the affected functions. The
+  largest gains are in `drop_read_counts()` and `position_from_variant_string()`
+  (now pure string operations), with smaller gains across most other functions.
+* Removed the unused `tidyr` dependency.