Dan Scott [Thu, 14 Jun 2012 19:16:51 +0000 (15:16 -0400)]
Avoid file extensions on scripts
Borrowing from the Debian policy, you might reimplement the same
functionality using a different language such as Perl or Python. So
avoid having a file extension in the first place. What are we, Windows
users?
Signed-off-by: Dan Scott <dan@coffeecode.net> Signed-off-by: Thomas Berezansky <tsbere@mvlc.org>
Dan Scott [Sun, 3 Jun 2012 03:25:53 +0000 (23:25 -0400)]
git2cl mangles the changelog; let's just use git's log
git2cl does bizarre things to line wrapping. The log it generates looks
deranged. Let's just use what git gives us; it's clean and easy to
parse, even if it doesn't adhere to the GNU changelog standards
(www.gnu.org/prep/standards/html_node/Style-of-Change-Logs.html).
Signed-off-by: Dan Scott <dan@coffeecode.net> Signed-off-by: Thomas Berezansky <tsbere@mvlc.org>
make_release.sh [-v VERSION] [-f PREV_BRANCH | -t | -b] [-F PREV_VERSION] [-n] [-p]
VERSION is auto-detected by default and is based on the currently checked out branch.
PREV_BRANCH is auto-detected by default but that isn't reliable. Include remote name!
PREV_VERSION Is auto-detected by default and is based on the PREV_BRANCH's name.
-n specifies that you don't want an upgrade script to be auto-generated.
-p specifies that this is a preview build.
-t turns on tag only mode.
-b turns on build only mode.
NOTE: -t and -b override PREV_BRANCH/PREV_VERSION, but -b overrides -t.
When building a release you should have git2cl in your PATH as well as
all packages needed for building i18n and staff clients.
Usage examples:
Tagging a major version root (no need for git2cl/staff client deps):
Release files are placed in a "release" folder one level up from your git
root directory.
Nothing is pushed to remote servers, but dojo and xulrunner files may be
downloaded. Dojo to install in the release tarball properly, xulrunner for
building the staff client and avoiding issues with xulrunner changes if
end-users build a custom staff client (say, for autoupdate purposes).
Signed-off-by: Thomas Berezansky <tsbere@mvlc.org> Signed-off-by: Dan Scott <dan@coffeecode.net>
Dan Scott [Sat, 16 Jun 2012 02:10:15 +0000 (22:10 -0400)]
TPAC: Add record detail navigation to page bottom
A usability rule of thumb suggests that navigation on long pages should
not require you to scroll back up to the top of the page from the
bottom; that gets old fast. We already do this on search results pages;
bring the same good practice to record details.
Signed-off-by: Dan Scott <dan@coffeecode.net> Signed-off-by: Dan Wells <dbw2@calvin.edu>
Security fix: Prevent login by deleted and barred users
An existing comment in the code suggested that we thought we were already
keeping barred users out. LP #1010671 brings up that deleted users were
not being kept out.
Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Revoking the UPDATE_MARC permission doesn't actually seem to prevent a
user from editing a record. Our use case is that we would like a user
to create new records but not edit existing MARC records.
Changing CREATE_MARC to UPDATE_MARC in OpenILS::Application::Cat's
biblio_record_replace_marc() method seems to fix the problem.
Signed-off-by: Jason Stephenson <jstephenson@mvlc.org> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Mike Rylander [Fri, 8 Jun 2012 13:26:25 +0000 (09:26 -0400)]
Remove unsafe thesaurus/control-set mapping
Evergreen 2.2 and beyond includes new Authority Control Set functionality
intended to allow the definition of locally defined authority control
headings. The primary mechanism for determining which Control Set is used
by any given authority record is to look at the thesaurus fixed field
(mnemonic: Subj), which currently must be uniquely linked to a Control Set.
Failing that (in the case of no thesaurus code at all), there is a backup
mechanism: if the thesaurus value for the authority record is missing or not
used by any configured Control Set then Evergreen will look at the first Main
Entry (1xx) field in the record, and find a control set which uses this field.
This is, obviously, not perfect, but for non-overlapping authority field
schemes, it works very well.
However, the seed data for Evergreen currently includes the "no attempt to
code" value of "|" in the set of thesauri that point to the LoC Control Set.
This value ("|") is also the default value for the Subj fixed field. This
means that the fallback mechanism is very likely never reached, and anything
short of fully specifying an appropriate thesaurus code for any non-LoC
authority record will run afoul of a similar outcome as the bib-data loss
situation described in bug #983487.
We can avoid this situation in most cases by /not/ tying the "|" thesaurus
value to the LoC control set.
This commit accomplishes this goal by allowing the Control Set for a thesaurus
to be NULL, removing the Control Set for "|" and removing the Control Set
recorded for records with "|" as the thesaurus value, thereby forcing Main
Entry inspection on any use of the authority record where Control Set is
required.
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Dan Wells [Tue, 12 Jun 2012 13:03:31 +0000 (09:03 -0400)]
Serial Note Sort Fix
Serial notes are currently sorting backwards from what is intended
(and what is expected based on other interfaces). This was caused
by adding a sort to the middle-layer code which resulted in an
unwanted double reverse.
[whitespace adjustment by LFW per request of DBS]
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
This accomplishes several things:
1 - Load the list of queues on page load
2 - Single-click access (rather than select and click) to queues
3 - Batch Deletion of Queues
Signed-off-by: Thomas Berezansky <tsbere@mvlc.org> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Dan Wells [Mon, 11 Jun 2012 15:18:08 +0000 (11:18 -0400)]
Show Vandelay Buckets in Manage Record Buckets
The 'Manage Record Buckets' view is currently filtered to show
only 'staff_client' type buckets. This expands the filter to
include 'vandelay_queue' buckets.
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Bill Erickson <berick@esilibrary.com>
Steven Callender [Thu, 19 Apr 2012 13:47:46 +0000 (09:47 -0400)]
Added a days_back parameter to authority_control_fields.pl in order to process bibs edited X amount of days ago.
The days_back parameter can be used to easily run the script through cron nightly and have it process all newly created or newly added bib records. Ideally it would be ran once a day at nighttime prior to midnight using the --days_back 0 option.
Signed-off-by: Steven Callender <stevecallender@esilibrary.com> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Dan Wells [Thu, 7 Jun 2012 14:32:59 +0000 (10:32 -0400)]
Fix broken non-RSS bookbag feeds
This commit addresses LP#1010036 by defining a stub 'description()'
method. Lack of such a method was causing internal server errors
for other kinds of feeds.
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Dan Scott [Wed, 6 Jun 2012 21:43:33 +0000 (17:43 -0400)]
Add TPAC translations
We're still working out our translation update process; new PO files can
fairly easily slip through the net. Fix that for TPAC, which is a pretty
big deal for 2.2.
Dan Scott [Wed, 6 Jun 2012 20:39:10 +0000 (16:39 -0400)]
Import updated translations from Launchpad
Lebbeous Fogle-Weekley originally imported these updated translations
from Launchpad. Dan Scott just weeded out a few metadata-only changes
to PO files that didn't need to be pushed to git.
Address LP #983487: Avoid clobbering bib records at authority merge
If you've set up a relationship between bib record and authority records
based on arbitrary sets of controlling and controlled fields, but your
authority records don't have the right Subj fixed field value
corresponding to a control set that defines the controlling and
controlled fields you're using, authority merge operations may wipe out
lots of bib data.
Yamil Suarez encountered this bug in testing, and in his case what he
needed to do was set the Subj fixed field in his authority records to
'A' to match his Song Title Index (see the launchpad bug referened
above).
Previously, you could not actually save the Subj fixed field in the MARC
editor (which showed HeadSubj instead of Subj for authority records, and
didn't work). Now you can.
Thanks to Mike Rylander for help in figuring out the above.
To provide additional protection against merging authority records when
they might not be linked with the right control set, the Manage
Authorities interface will now also show you the linked control set for
any given records.
The implementation of that last bit incidentally meant making sure flesh
and flesh_fields get passed through to PermaCrud for retrieve() calls via
the Javascript openils.PermaCrud wrapper.
Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com> Signed-off-by: Bill Erickson <berick@esilibrary.com>
It appears as though a copy-and-paste error introduced duplicate keys in
circ.properties, which in turn makes newer versions of the Translate
Toolkit unhappy and results in empty translated files. Tracing through
the pertinent code in xul/staff_client/server/circ/util.js, it looks
like the desired key was for "not_found" for the second "cataloging" key
- so change that accordingly.
Signed-off-by: Dan Scott <dan@coffeecode.net> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Dan Scott [Sun, 3 Jun 2012 05:09:26 +0000 (01:09 -0400)]
Improve Fedora prereqs and instructions
Fedora 16 and 17 need the makensis package.
As well, make the Makefile.install target "fedora" to match the OpenSRF
target, and adjust the docs accordingly. This should lead to fewer
changes in the future as Fedora 18, 19, etc roll out.
Bill Erickson [Thu, 3 May 2012 20:36:38 +0000 (16:36 -0400)]
TPAC: Prevent unitialized variable warnings in My Account (repair)
The logic in the second IF test was slightly altered by the fix. An
undef value for username_unlimit should result in executing the block.
Since org setting values are stored as real boolean values, I simplified
the test to a simple if(!username_unlimit) test.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Dan Scott <dan@coffeecode.net>
Bill Erickson [Fri, 23 Mar 2012 19:17:50 +0000 (15:17 -0400)]
ACQ guarantee no PO re-activation
Prevent the possibility of double PO activation by preventing it within
the API. There have been reports in the wild of double-activation, even
though the UI is supposed to prevent it.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Michael Peters <mrpeters@library.in.gov> Signed-off-by: Dan Scott <dan@coffeecode.net>
Fix broken menu entries under "Other -> Booking" in patron interface
This issue was reported to me by Steve Callender. The menu entries
under "Other -> Booking" had all stopped working.
There's a catalog of URL constants under window.urls sometimes, and
formerly (or other times?) at xulG.urls for XUL-based staff client
interfaces. There's a new-tab-opening method that needed to be more
flexible when looking for that catalog.
Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com> Signed-off-by: Bill Erickson <berick@esilibrary.com>
Bill Erickson [Fri, 1 Jun 2012 14:28:28 +0000 (10:28 -0400)]
Override arguments parameter repairs
In some cases, the override arguments were passed as a paramater after
an @array or %hash argument. In such cases, any params following are
clumped into the array/hash and the trailing parameters get no values.
Let's explicitly push the $oargs into the preceeding structures.
In the case of Circulate.pm:new, no $oargs argument is needed, as all keys
within the %args hash are tacked onto the circulator object.
Also, for consistency, this change affects one of the hold API calls by
moving $oargs into $params even when not strictly necessary.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Galen Charlton [Fri, 1 Jun 2012 03:31:11 +0000 (23:31 -0400)]
lp#1007248: tweak order in which pgt rows are renumbered
Need to renumber from highest to lowest. Because the primary
key constraint is not deferred, doing the renumbering in a single
statement will fail if there is overlap between the set of
existing pgt.id values and the set of {pgt.id + 100}.
Signed-off-by: Galen Charlton <gmc@esilibrary.com> Signed-off-by: Bill Erickson <berick@esilibrary.com>
Bill Erickson [Fri, 25 May 2012 18:41:18 +0000 (14:41 -0400)]
TPAC bookbag CGI param changed to bbid
To avoid propagating a generic URL parameter like 'id', which can be
picked up by other pages and cause breakage, use 'bbid' instead. Also,
clear the 'bbid' param from the my-account tabs.
While we're in there, silence an uninitialized string warning.
Signed-off-by: Bill Erickson <berick@esilibrary.com>
Jeff Davis [Wed, 11 Apr 2012 16:04:19 +0000 (12:04 -0400)]
LP#856708 - Scrolling/div size issues in action trigger interface
The event definition content pane does not properly accommodate its contents.
In the Event Definitions tab, the list of event definitions exceeds the height of the
pane, but you can't scroll all the way down to the bottom of the list -- some of the
content gets cut off. Likewise, when editing an event definition, you can't scroll
all the way to the bottom of the page; the Cancel and Save buttons are cut off, and
you have to use the Tab key to move the focus down to those buttons, which forces
the pane to scroll to the bottom so that they're visible.
Signed-off-by: Jeff Davis <jdavis@sitka.bclibraries.ca> Signed-off-by: Michael Peters <mrpeters@library.in.gov> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Dan Scott [Thu, 12 Apr 2012 01:56:56 +0000 (21:56 -0400)]
Group acquisition OU settings in a new "Acquisitions" group
As seen in LP867465, the current grouping of acquisition OU settings
under the generic "lib" group leads to confusion about which
context "default shelf location" and the like apply to. Creating a new
"Acquisitions" group will help clarify that current ambiguity.
Signed-off-by: Dan Scott <dan@coffeecode.net> Signed-off-by: Bill Erickson <berick@esilibrary.com>
Acq: Use scope to constrain choice of lineitem alerts
This commit addresses this launchpad bug:
https://bugs.launchpad.net/evergreen/+bug/996033
The controlled vocabulary of acq lineitem alert text is set up in an
interface labeled "Line Item Alerts" in the staff client admin menu. In
the PO and selection list interfaces (where alerts can be applied), we
will limit the user to selecting alerts from those owned at the user's
workstation org unit (or its ancestors).
When we display applied alerts, we can also display the shortname of the
org unit at which the alert text is defined (the owning_lib column).
This should help clarify what's going on with alerts that have already
been applied using alert text from another org unit.
[Amended since original reference on Launchpad to also show the
owning_lib's shortname when confirming alerts at receive time.]
Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com> Signed-off-by: Bill Erickson <berick@esilibrary.com>
Jason Etheridge [Tue, 10 Apr 2012 19:35:31 +0000 (15:35 -0400)]
menu entry Circulation->Clear Shelf-Expired Holds
This spawns the Browse Hold Shelf interface but automatically triggers the View
Shelf-Expired Holds and Clear these Holds options.
Signed-off-by: Jason Etheridge <jason@esilibrary.com> Signed-off-by: Michael Peters <mrpeters@library.in.gov> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Tue, 22 May 2012 21:40:04 +0000 (17:40 -0400)]
TPAC auto-override permissable patron hold fail events
This adds a new org unit setting which tells the tpac to automatically
override hold placement failure events in cases where the patron has the
permission to do so. The goal is to skip the confirmation step
currently required by patrons when overridding a tpac hold.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
This adds a new parameter to backend override functions. In most cases the
system will act like a hash with all set to 1 was passed in when nothing was
passed in by the caller.
The new parameter should be a hash containing zero or more of the following
options:
all - If true (1) override all events
events - Should be an array of event textcodes to override
For hold placement the "all" option is ignored for possibility checks, but
is checked for other events like HOLD_EXISTS.
Also included for hold placement is to always run possibility checks for
the test and create batch method.
Overall, the default is to act like things did previously.
Signed-off-by: Thomas Berezansky <tsbere@mvlc.org> Signed-off-by: Mike Rylander <mrylander@gmail.com>
New pull list interface taking advantage of flattener for speed,
and advanced sorting. For now, access it by the "Simplifed Pull List"
button along the bottom edge of the existing holds pull list interface
(but I think when/if this thing is widely accepted, it should replace
the existing interface outright).
With thanks to Mike Peters for testing an early version. Now including
some updates requested by Thomas Berezansky. Specifically, the
queue_position column and its relatives fthat come from the same
calculation were removed, as they're [very] expensive in computing time to
produce and generally useless to pull lists.
One exception to the characterization of those fields as "useless" is the
"number of potential copies" column, which we should add back later
assuming we can find a fast way to do it rather than the previous ways,
which were slow.
Thanks to Bill Erickson for helping fix my issues in making auto-generated
columns coöperate with the column picker (his changes are squashed into
this). I think it's finally right.
Now with release notes.
Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Fri, 18 May 2012 17:23:53 +0000 (13:23 -0400)]
tpac added content default type improvements
* In the pre-JS state, only choose a default display type when the type
is known to have data.
* If no default type is chosen for the main tab link, because no data
was available to the template, set the default from the first added
content type determined to be available via JS.
* Move added content JS to separate template file for easier override.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Tue, 15 May 2012 14:11:37 +0000 (10:11 -0400)]
TPAC added content data retrieval sanity check
On the backend, avoid fetching added content data when initial request
returns 404. This prevents edge case where a specific type may be
selected (e.g. via bookmark) even though no data exists for the type.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Mon, 14 May 2012 19:22:08 +0000 (15:22 -0400)]
TPAC added content integration
https://bugs.launchpad.net/evergreen/+bug/984963
This adds a new tab on the detail page called Additional Content
(suggestions welcome). When the tab is expanded, available content is
presented to the user via a series of sub-tabs.
At the start of loading the record detail page, kick off a series of
asynchronous HTTP HEAD requests, one per type of added content. At the
end of context loading, read the results of the HTTP requests for any
that have completed. If the status for a type is 200, the type is
marked as available. If it's not 200 (usually 404) it's marked as not
available. Otherwise, it's marked as unknown.
In the template, available content produces a link the user can click
to view the content. Non-available content produces no links. Unknown
content produces a link that may lead to content or no content when
JS/dojo is disabled. When dojo is enabled, unknown content sends a
series of JS queries to determine the state of the content
asynchronously and results in displayed links for any type that is later
determined to have content.
TODO: when the user expands the Additional Content tab, no type is
chosen by default. A default (available) type needs to be selected on
initial display.
TODO: better styling, particularly in the sub-tabs.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Tue, 22 May 2012 20:22:14 +0000 (16:22 -0400)]
TPAC: recover the ability to override hold placement failures
There is a certain class of hold failure events (e.g. HOLD_EXISTS,
HOLD_ITEM_CHECKED_OUT) that can be overridden by patrons when they have
the correct override permission. This change recovers that ability,
which was recently removed with 99e8fc893a.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
The set of filter groups available to the TPAC are those defined at the
search org unit and its ancestors. If multiple filters have the same
code name, the filter owned closest to the context org unit is used.
This allows for global filters with local overrides.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Wed, 18 Apr 2012 20:15:46 +0000 (16:15 -0400)]
tpac : search_HINT cache fixes
When searching for multiple values like a list of IDs, generate the
cache key from the sorted JSON form of the search value instead of the
bare ARRAY ref.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Mike Rylander [Wed, 18 Apr 2012 17:24:24 +0000 (13:24 -0400)]
Teach QueryParser about search filter callbacks, and if a callback returns a value, parse it at the current recursion level
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Thu, 23 Feb 2012 19:28:50 +0000 (14:28 -0500)]
Search Filter Groups and Generic Queries : DB / IDL
Adds 3 new tables. The first is a general purpose container for
query-parser queries. These can be used by other entities for storing
saved searches, saved search filters, etc. Each contains a query string
and a label.
The two other tables are for creating org unit saved search filter
groups. A filter group is a named set of search queries, treated as
filters, collected to create a high-level search filter. Filter group
entries use search_query's under the covers, so the content of the
filter can be any valid query string, including other filters or even
other filter groups.
For example, you could create an "Audience" filter group with entries
like, "Children" => "audience(a, b) locations(3,4,5)" to filter on copy
location in addition to the MARC audience data.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Mon, 21 May 2012 14:32:56 +0000 (10:32 -0400)]
Query Parser merged copy / non-dynamic filters
When more than one copy-level filter (e.g. locations()) is used in a
query-parser query, all but the first are ignored. The goal of this
work is to compress multiple copy-level filters into a single filter
that represents the full collection. Wherever possible, boolean
structures will be honored, to the extent possible for post-search
copy-level filters.
( concerto locations(3,4,5) ) && ( piano locations(3,5,7) )
==> concerto piano locations(3,5)
Note, in the last 2 examples, the final query does not exactly match the
original. This is because copy-level filters are applied after the
initial search and cannot be executed as part of the nested query. The
best we can do is create a representation of the final copy-level
filter, based on the nesting, and apply it to the final result of the
nested search.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Tue, 15 May 2012 22:00:30 +0000 (18:00 -0400)]
Query Parser nested dynamic filters
Dynamic filters (from metabib.record_attr) are now parsed along with
search strings and facets as Query Parser descends the parse tree. The
WHERE clause for the query as a whole is now a combination of the main
WHERE and the ON clause for JOINing to metabib.record_attr. This gives
us the ability to create complex boolean nested queries using
SVF/record_attr filters, unlike before where nested filters were either
ignored or always treated as global filters.
Dan Scott [Mon, 21 May 2012 19:09:59 +0000 (15:09 -0400)]
Use spaces instead of tabs in QueryParser.pm
Copying swathes of SQL from logs into psql would often trigger the
autocompletion that psql offers, garbling a query awfully. Define the
local $spc var as two spaces and use that instead, leaving open the
option for reverting to tabs or anything else people prefer.
Also remove trailing whitespace, as this whole thing is about space.
Signed-off-by: Dan Scott <dan@coffeecode.net> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Mike Rylander [Thu, 29 Dec 2011 16:36:20 +0000 (11:36 -0500)]
abstract query representations from QueryParser
When you perform a search, you now also get back an abstract
representation of your search query. fts-replacement.pl is a test
script that shows you things that QueryParser does, including the new
abstract query feature, and it also reconstitutes a string from the
abstract query to demonstrate the correctness of that query.
[Also]
Make last-wins true for facets, filters and modifiers -- arg!; Parameterized values win in the staged-search QP wrapper
Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com> Signed-off-by: Dan Scott <dan@coffeecode.net>
Dan Scott [Mon, 21 May 2012 18:46:53 +0000 (14:46 -0400)]
QueryParser: comma-delimit multiple WITH clauses
Bill Erickson's fix for my breakage of QueryParser.pm for the speedup
via the WITH common table expression usage inadvertently left out
the commas required to delimit multiple WITH clauses, resulting in SQL
query errors.
This fixes that oversight. Note that the fts-replacement.pl script came
in quite handy for testing this; recommended.
Signed-off-by: Dan Scott <dan@coffeecode.net> Signed-off-by: Bill Erickson <berick@esilibrary.com>
Jason Stephenson [Sat, 19 May 2012 02:51:14 +0000 (22:51 -0400)]
Add libnet-https-any-perl as precise deb requirement in Makefile.install.
Business::OnlinePayment::PayPal requires it, and it does not always want
to install from CPAN. It says it require either Net-SSLeay or
Crypt-SSLeay but the CPAN package never wants to install unless they
both are installed. The deb package for Ubuntu seems to do the right
thing.
Signed-off-by: Jason Stephenson <jason@sigio.com> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
Bill Erickson [Fri, 18 May 2012 20:42:19 +0000 (16:42 -0400)]
QP support for nested WITH construction
Nested query plans build their own tsquery chunks which have to be
individually JOINed to the final query. To do this, each tsquery SELECT
gets its own entry in the top-level WITH, complete with alias for
joining.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Dan Scott [Mon, 21 May 2012 03:02:09 +0000 (23:02 -0400)]
Further integration: link README to server install docs
We will now have the INSTALL file linking to the README file which in
turn leads to the docs/installation/server_installation.text file. But
it all works!
Dan Scott [Mon, 21 May 2012 02:51:52 +0000 (22:51 -0400)]
Merge the Evergreen docs into the code repository
Rather than maintaining two separate repositories, with dual maintenance
of what are essentially single-sourced files in the release notes and
the install instructions, make better use of the docs/ subfolder of the
Evergreen repository. If this crazy scheme works, DIG can commit to the
docs/ subfolder, and devs can commit there too.
Dan Scott [Wed, 16 May 2012 04:29:11 +0000 (00:29 -0400)]
Optimize full text search with a CTE
The tsquery used directly in a JOIN was resulting in bad
plan with lots of nested loops and operations. Using a CTE
to define the query improves the plan.
Also, instead of ARRAY_AGG(DISTINCT foo), just return the
whole thing in an array. Not great for dupes, but we can
handle those inside search.query_parser_fts if we really
need to.
Signed-off-by: Dan Scott <dscott@laurentian.ca> Signed-off-by: Mike Rylander <mrylander@gmail.com>
Bill Erickson [Wed, 16 May 2012 19:55:23 +0000 (15:55 -0400)]
Vandelay authority existing queue search repair
When searching for an existing authority queue, search for an authority
queue, not a bib queue. Apart from simply not working, the query will
fail with an "invalid input value for enum vandelay.bib_queue_queue_type"
and result in a client-side error.
https://bugs.launchpad.net/evergreen/+bug/989391
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Lebbeous Fogle-Weekley <lebbeous@esilibrary.com>
projects / working / Evergreen.git / log