aboutsummaryrefslogtreecommitdiffstats
path: root/notes
diff options
context:
space:
mode:
authorBryan Newbold <bnewbold@robocracy.org>2021-11-09 15:46:20 -0800
committerBryan Newbold <bnewbold@robocracy.org>2021-11-09 15:46:20 -0800
commit996b2e2084c1798126bd91dd950c063982398bec (patch)
tree2c4a9eef6432158088fea255db9e8b7b098a371d /notes
parenta246b5a54dac6b29a30e90265d64a3c4332902e5 (diff)
downloadfatcat-996b2e2084c1798126bd91dd950c063982398bec.tar.gz
fatcat-996b2e2084c1798126bd91dd950c063982398bec.zip
more iteration on short wayback timestamp cleanup
Diffstat (limited to 'notes')
-rw-r--r--notes/cleanups/scripts/fetch_full_cdx_ts.py2
-rw-r--r--notes/cleanups/wayback_timestamps.md129
2 files changed, 128 insertions, 3 deletions
diff --git a/notes/cleanups/scripts/fetch_full_cdx_ts.py b/notes/cleanups/scripts/fetch_full_cdx_ts.py
index 6f67c7e1..d5b0c476 100644
--- a/notes/cleanups/scripts/fetch_full_cdx_ts.py
+++ b/notes/cleanups/scripts/fetch_full_cdx_ts.py
@@ -137,7 +137,7 @@ def process_file(fe, session) -> dict:
if short in full_urls:
continue
- if original_url in self_urls:
+ if original_url in self_urls and ts in self_urls[original_url]:
full_urls[short] = self_urls[original_url]
status = "success-self"
continue
diff --git a/notes/cleanups/wayback_timestamps.md b/notes/cleanups/wayback_timestamps.md
index 81785992..85e5f94f 100644
--- a/notes/cleanups/wayback_timestamps.md
+++ b/notes/cleanups/wayback_timestamps.md
@@ -61,9 +61,10 @@ Yes, 4-digit is a popular pattern as well, need to handle those:
# 111M 0:13:22 [ 139k/s]
zcat files_20211007_moreshortts.json.gz | wc -l
+ # 9,958,854
zcat files_20211007_moreshortts.json.gz | shuf -n10000 > files_20211007_moreshortts.10k_sample.json
- # 9,958,854
+
## Fetch Complete URL
@@ -114,6 +115,14 @@ Again with the broader set:
6175 "success-db"
3035 "success-self"
+While running a larger batch, got a CDX API error:
+
+ requests.exceptions.HTTPError: 403 Client Error: Forbidden for url: https://web.archive.org/cdx/search/cdx?url=https%3A%2F%2Fwww.psychologytoday.com%2Ffiles%2Fu47%2FHenry_et_al.pdf&from=2017&to=2017&matchType=exact&output=json&limit=20
+
+ org.archive.util.io.RuntimeIOException: org.archive.wayback.exception.AdministrativeAccessControlException: Blocked Site Error
+
+So maybe need to use credentials after all.
+
## Cleanup Process
@@ -128,11 +137,13 @@ It looks like the rel swap is already implemented in `generic_file_cleanups()`.
From sampling it seems like the mimetype issue is pretty small, so not going to
bite that off now. The "bogus file" issue requires thought, so also skipping.
-## Commands
+
+## Commands (old)
Running with 8x parallelism to not break things; expecting some errors along
the way, may need to add handlers for connection errors etc:
+ # OLD SNAPSHOT
zcat files_20211007_moreshortts.json.gz \
| parallel -j8 --linebuffer --round-robin --pipe ./fetch_full_cdx_ts.py \
| pv -l \
@@ -140,3 +151,117 @@ the way, may need to add handlers for connection errors etc:
> files_20211007_moreshortts.fetched.json.gz
At 300 records/sec, this should take around 9-10 hours to process.
+
+
+
+## Prep Again (2021-11-09)
+
+After fixing "sort" issue and re-dumping file entities (2021-11-05 snapshot).
+
+Filter again:
+
+ # note: in the future use pigz instead of gzip here
+ zcat file_export.json.gz \
+ | pv -l \
+ | rg 'web.archive.org/web/\d{4,12}/' \
+ | gzip \
+ > files_20211105_moreshortts.json.gz
+ # 112M 0:13:27 [ 138k/s]
+
+ zcat files_20211105_moreshortts.json.gz | wc -l
+ # 9,958,854
+ # good, exact same number as previous snapshot
+
+ zcat files_20211105_moreshortts.json.gz | shuf -n10000 > files_20211105_moreshortts.10k_sample.json
+ # done
+
+ cat files_20211105_moreshortts.10k_sample.json \
+ | ./fetch_full_cdx_ts.py \
+ | pv -l \
+ > files_20211105_moreshortts.10k_sample.fetched.json
+ # 10.0k 0:03:36 [46.3 /s]
+
+ cat files_20211105_moreshortts.10k_sample.fetched.json | jq .status | sort | uniq -c
+ 13 "fail-not-found"
+ 774 "success-api"
+ 6193 "success-db"
+ 3020 "success-self"
+
+After tweaking `success-self` logic:
+
+ 13 "fail-not-found"
+ 859 "success-api"
+ 6229 "success-db"
+ 2899 "success-self"
+
+
+## Testing in QA
+
+Copied `sample_out.json` to fatcat QA instance and renamed as `files_20211007_moreshortts.10k_sample.fetched.json`
+
+ # OLD ATTEMPT
+ export FATCAT_API_AUTH_TOKEN=[...]
+ head -n10 /srv/fatcat/datasets/files_20211007_moreshortts.10k_sample.fetched.json \
+ | python -m fatcat_tools.cleanups.file_short_wayback_ts -
+
+Ran in to issues, iterated above.
+
+Trying again with updated script and sample file:
+
+ export FATCAT_AUTH_WORKER_CLEANUP=[...]
+
+ head -n10 /srv/fatcat/datasets/files_20211105_moreshortts.10k_sample.fetched.json \
+ | python -m fatcat_tools.cleanups.file_short_wayback_ts -
+ # Counter({'total': 10, 'update': 10, 'skip': 0, 'insert': 0, 'exists': 0})
+
+Manually inspected and these look good. Trying some repeats and larger batched:
+
+ head -n10 /srv/fatcat/datasets/files_20211105_moreshortts.10k_sample.fetched.json \
+ | python -m fatcat_tools.cleanups.file_short_wayback_ts -
+ # Counter({'total': 10, 'skip-revision-changed': 10, 'skip': 0, 'insert': 0, 'update': 0, 'exists': 0})
+
+ head -n1000 /srv/fatcat/datasets/files_20211105_moreshortts.10k_sample.fetched.json \
+ | python -m fatcat_tools.cleanups.file_short_wayback_ts -
+
+ [...]
+ bad replacement URL: partial_ts=201807271139 original=http://www.scielo.br/pdf/qn/v20n1/4918.pdf fix_url=https://web.archive.org/web/20170819080342/http://www.scielo.br/pdf/qn/v20n1/4918.pdf
+ bad replacement URL: partial_ts=201904270207 original=https://www.matec-conferences.org/articles/matecconf/pdf/2018/62/matecconf_iccoee2018_03008.pdf fix_url=https://web.archive.org/web/20190501060839/https://www.matec-conferences.org/articles/matecconf/pdf/2018/62/matecconf_iccoee2018_03008.pdf
+ bad replacement URL: partial_ts=201905011445 original=https://cdn.intechopen.com/pdfs/5886.pdf fix_url=https://web.archive.org/web/20190502203832/https://cdn.intechopen.com/pdfs/5886.pdf
+ [...]
+
+ # Counter({'total': 1000, 'update': 969, 'skip': 19, 'skip-bad-replacement': 18, 'skip-revision-changed': 10, 'skip-bad-wayback-timestamp': 2, 'skip-status': 1, 'insert': 0, 'exists': 0})
+
+
+It looks like these "bad replacement URLs" are due to timestamp mismatches. Eg, the partial timestamp is not part of the final timestamp.
+
+Tweaked fetch script and re-ran:
+
+ # Counter({'total': 1000, 'skip-revision-changed': 979, 'update': 18, 'skip-bad-wayback-timestamp': 2, 'skip': 1, 'skip-status': 1, 'insert': 0, 'exists': 0})
+
+Cool. Sort of curious what the deal is with those `skip-bad-wayback-timestamp`.
+
+Run the rest through:
+
+ cat /srv/fatcat/datasets/files_20211105_moreshortts.10k_sample.fetched.json \
+ | python -m fatcat_tools.cleanups.file_short_wayback_ts -
+ # Counter({'total': 10000, 'update': 8976, 'skip-revision-changed': 997, 'skip-bad-wayback-timestamp': 14, 'skip': 13, 'skip-status': 13, 'insert': 0, 'exists': 0})
+
+Should tweak batch size to 100 (vs. 50).
+
+How to parallelize import:
+
+ # from within pipenv
+ cat /srv/fatcat/datasets/files_20211105_moreshortts.10k_sample.fetched.json \
+ | parallel -j8 --linebuffer --round-robin --pipe python -m fatcat_tools.cleanups.file_short_wayback_ts -
+
+
+## Full Batch Commands
+
+Running in bulk again:
+
+ zcat files_20211105_moreshortts.json.gz \
+ | parallel -j8 --linebuffer --round-robin --pipe ./fetch_full_cdx_ts.py \
+ | pv -l \
+ | gzip \
+ > files_20211105_moreshortts.fetched.json.gz
+