These were never used in the function. Where they are used is in the
pkg_details.php template, so move them closer to their actual usage so
as not to confuse poor programmers such as myself.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
This prevents needless calls to db_connect() if we already have a
reference to a handle we can pass into a function. Although the current
underlying implementation using mysql_connect() will return the same
connection if all parameters are the same, this might not be true if we
switch to a more modern (e.g. PDO) interface.
In the face of safe transactions, it is extremely important all actions
are being taken over the same connection to the database.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
This allows for merging comments and votes of deleted packages into
another one which is useful if a package needs to be renamed.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Signed-off-by: Dan McGee <dan@archlinux.org>
Describe what this function actually does: Return the ID of a package
with a given name and return NULL if such a package doesn't exist.
The function name is chosen in a fashion similar to other functions from
"pkgfuncs.inc.php" (pkgname_from_id(), pkgnotify_from_sid(), ...).
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
This implements the following scheme:
* /packages/cower/ --> /packages/co/cower/
* /packages/j/ --> /packages/j/j/
* /packages/zqy/ --> /packages/zq/zqy/
We take up to the first two characters of each package name as a
intermediate subdirectory, and then the full package name lives
underneath that. Shorter named packages live in a single letter
directory.
Why, you ask? Well because earlier today the AUR hit 32,000 entries in
the unsupported/ directory, making new package uploads impossible. While
some might argue we shouldn't have so many damn packages in the repos,
we should be able to handle this case.
Why two characters instead of one? Our two biggest two-char groups, 'pe'
and 'py', both start with 'p', and have nearly 2000 packages each. Go
Python and Perl.
Still needed is a "move the existing data" script, as well as a set of
rewrite rules for those wishing to preserve backward compatible URLs for
any helper programs doing the wrong thing and relying on them.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
We shouldn't require this as it is a new config parameter and it causes
PHP warnings to be spewed everywhere.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
There was too much voodoo going on in new_sid(). Just use uniqid() with
a random seed and the optional entropy parameter to generate MD5 input.
Use the remote IP address as a salt to reduce the chance of two clients
getting the same ID if they login at exactly the same time.
Thanks-to: Florian Pritz <bluewind@xinu.at>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Use PHP's filter extension to validate mail addresses. The filter
extension is enabled by default as of PHP 5.2.0.
Thanks-to: Florian Pritz <bluewind@xinu.at>
Thanks-to: Cédric Girard <girard.cedric@gmail.com>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
We were doing some silly things here with an "ORDER BY Name, CategoryID"
clause, due to the fact that Name is unique, and thus any additional
ordering after Name will have no effect. Of course, the dumb as a box of
rocks MySQL query optimizer doesn't realize this, leading to full table
scans every time of ~30000 packages instead of using index scans for the
ordering and only retrieving the first 50 rows.
The biggest change is noted in the default sort order as it cuts down
the columns to one, but we can remove the redundant sort from other
orderings as well, even though those will still not be able to use an
index-driven query plan.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
The use of header() to redirect after previous output was a design flaw.
Our only luck here was PHP's "output_buffering" config variable
defaulting to 4096, which kind of hid the bug for a long time.
Fixes FS#24580.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Provie a mechanism to specify cache type from NONE, APC, or MEMCACHE
based on a config variable.
If MEMCACHE type is selected, a list of servers can be specified to
provide multiserver support. Note that php-memcaced is required for
MEMCACHE support.
Lukas: Minor commenting style fixes.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
- move apc cache code to aur.inc (centralize)
- refactor the apc usage in stats.inc to utilize new code in aur.inc
Lukas: Small commenting style and spelling changes.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
after reading the discussion about the submitter field, i looked again
at my previous patch and realized that it was not very good at all.
hopefully this simplified version will be more acceptable.
the justification for this patch is that many third-party tools will
want to know how new a package is, or when it last changed. adding the
submitted and modified timestamps will allow users to get all the
relevant package data in a single query.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Add etag and if-none-match conditional get support. This will allow
'smart client' to save network bandwidth, as they can save the etag hash
value for queries and test it later. Still an http request because this
patch specifically sets a cache lifetime of zero, and must-revalidate.
The benefit here is bandwidth savings. Caching based on expires headers
would likely be counter productive, as the api data can change rather
quickly...but etag is a nice compromise, and could be quite beneficial
for bandwidth recution in some scenarios.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
the query was being performed when $id was not set, resulting in an
invalid sql query being performed.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
make the sql query form consistent in usage by cleaning up instances
where db_query's result was not inspected before attempting to fetch row
data from the handle
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
add a hook to db_query to log all sql queries when SQL_DEBUG is set
Additionally, provide better logging for sql error situations (provide
backtrace as well as error message).
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Ensure we are not quoting these values in any of our SQL queries.
Thanks-to: elij <elij.mx@gmail.com>
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
The majority of "real world" info requests [1] come in hefty batches. We
would be better served to handle these in one request rather than
multiple by allowing AUR clients to send multiple arguments.
This enables things like this to work:
http://aur.test/rpc.php?type=multiinfo&arg[]=cups-xerox&arg[]=cups-mc2430dl&arg[]=10673
Note to RPC users: unfortunately due to the asinine design of PHP, you
unfortunately have to use the 'arg[]' syntax if you want more than one
query argument, or you will only get the package satisfying the last arg
you pass.
[1] Rough data from April 11, 2011, with a total hit count of 1,109,163:
12 /login.php
13 /rpc.php?type=sarch
15 /rpc.php?type=msearch
16 /pingserver.php
16 /rpc.php
22 /logout.php
163 /passreset.php
335 /account.php
530 /pkgsubmit.php
916 /rss2.php
3838 /index.php
6752 /rss.php
9699 /
42478 /rpc.php?type=search
184737 /packages.php
681725 /rpc.php?type=info
That means a whopping 61.5% of our requests were for info over the RPC
interface; package pages are a distant second at only 16.7%.
Lukas: Introduce "multiinfo" query instead of extending "info" (for the
sake of backward compatibility).
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Do the implode as the same but separate step each time, and remove
indentation where no other query has it.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
* Mark things static in the class rather than use a constructor every
single invocation of the service.
* Don't call mysql_real_escape_string() before we even have a database
connection, and don't do work in the database if we don't need to.
* Formatting consistency fixups in a few places.
* Add new process_query() helper function; use this instead of
copy-pasted code in all of the RPC method calls.
* Remove the escaping code meant to fix FS#15526, introduced in commit
4d1eb4dd7a. It broke more than it solved, only fixed the output in
one of three RPC calls (and who knows what the web interface then also
does), and proper encoding should be done at the database level rather
than up here.
Lukas: Add special case for "info" queries to process_query() (return a
single result instead of an array of results here).
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
You need this enabled for the AUR, period. No need for this BS.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Ensures "$name" is always initialized, even if the package doesn't
exist. Fixes PHP warnings appearing when retrieving package details of a
package with an invalid ID.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
We removed the code depending on this a long time ago - drop it and add
some note to "UPGRADING".
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Instead, we just store dependencies directly in the PackageDepends
table. Since we don't use this info anywhere besides the package details
page, there is little value in precalculating what is in the AUR vs.
what is not.
An upgrade path is provided via several SQL statements in the UPGRADING
document. There should be no user-visible change from this, but the DB
schema gets a bit more sane and we no longer have loads of junk packages
in our tables that are never shown to the end user. This should also
help the MySQL query planner in several cases as we no longer have to be
careful to exclude dummy packages on every query.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Set it equal to the SubmittedTS field, which will be our indication the
package is new when we show the logo on the front page of the AUR.
This results in the ability to remove the use of the unindexable
GREATEST() function from the AUR code everywhere we had to use it before
to handle the 0 timestamp case.
Note that there is no race condition here in calling UNIX_TIMESTAMP()
twice- it always returns the time at the beginning of statment
execution:
mysql> select unix_timestamp(), sleep(2), unix_timestamp();
+------------------+----------+------------------+
| unix_timestamp() | sleep(2) | unix_timestamp() |
+------------------+----------+------------------+
| 1300851746 | 0 | 1300851746 |
+------------------+----------+------------------+
1 row in set (2.00 sec)
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
25 is woefully small for the number of packages many searches can
return, and with 28000+ packages in AUR, it makes sense to show a lot
more per page by default.
The new choices of (50, 100, 250) happen to match those from the main
site.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Regression introduced in commit ef8fab0c. Removed the strict check again
instead of wrapping it in __() to ensure search URLs are language
independent.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
Fix a lot of invalid XHTML in the templates and actions. There might
still be some legacy code left, but this should cover most of it.
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
This removes the need for SQL_CALC_FOUND_ROWS which can really slow down
queries in a lot of cases. The COUNT(*) query we end up performing can
reuse a lot of the original clauses from our primary query, but we can
really slim it up by omitting some joins and the sorting/limiting
clauses.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
No functional change here; we should be rebuilding the same query at the
end of the process.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
And use implode() instead of some looping/first time logic.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
With our use of strtotime() in stats.inc, we are "required" to do so or at
least the emitted warning tells us it is a good idea.
Signed-off-by: Dan McGee <dan@archlinux.org>
Signed-off-by: Lukas Fleischer <archlinux@cryptocrack.de>
