summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorBryan Newbold <bnewbold@robocracy.org>2021-11-09 18:14:58 -0800
committerBryan Newbold <bnewbold@robocracy.org>2021-11-09 18:49:46 -0800
commit23fd36a3e8505c1ed6d13367a3fb62a8bf2242d7 (patch)
treea42efe1b89edb8c8205b9bdd480615510f7814ae
parent1024e688bb12d64648ceb638daf049d508f87561 (diff)
downloadfatcat-23fd36a3e8505c1ed6d13367a3fb62a8bf2242d7.tar.gz
fatcat-23fd36a3e8505c1ed6d13367a3fb62a8bf2242d7.zip
add notes about 'double slash in DOI' issue
-rw-r--r--notes/cleanups/double_slash_dois.md46
1 files changed, 46 insertions, 0 deletions
diff --git a/notes/cleanups/double_slash_dois.md b/notes/cleanups/double_slash_dois.md
new file mode 100644
index 00000000..d4e9ded6
--- /dev/null
+++ b/notes/cleanups/double_slash_dois.md
@@ -0,0 +1,46 @@
+
+Relevant github issue: https://github.com/internetarchive/fatcat/issues/48
+
+
+## Investigate
+
+At least some of these DOIs actually seem valid, like
+`10.1026//1616-1041.3.2.86`. So shouldn't be re-writing them!
+
+ zcat release_extid.tsv.gz \
+ | cut -f1,3 \
+ | rg '\t10\.\d+//' \
+ | wc -l
+ # 59,904
+
+ zcat release_extid.tsv.gz \
+ | cut -f1,3 \
+ | rg '\t10\.\d+//' \
+ | pv -l \
+ > doubleslash_dois.tsv
+
+Which prefixes have the most double slashes?
+
+ cat doubleslash_dois.tsv | cut -f2 | cut -d/ -f1 | sort | uniq -c | sort -nr | head
+ 51220 10.1037
+ 2187 10.1026
+ 1316 10.1024
+ 826 10.1027
+ 823 10.14505
+ 443 10.17010
+ 186 10.46925
+ 163 10.37473
+ 122 10.18376
+ 118 10.29392
+ [...]
+
+All of the 10.1037 DOIs seem to be registered with Crossref, and at least some
+have redirects to the not-with-double-slash versions. Not all doi.org lookups
+include a redirect.
+
+I think the "correct thing to do" here is to add special-case handling for the
+pubmed and crossref importers, and in any other case allow double slashes.
+
+Not clear that there are any specific cleanups to be done for now. A broader
+"verify that DOIs are actually valid" push and cleanup would make sense; if
+that happens checking for mangled double-slash DOIs would make sense.