As I see, the problem is that ajax request (may be not ajax only) didn't
wait for closing session file by another one and read/save data from/to it
at any time! Lock did not work on my Win7 because , as I think, for Win
it's one process.
Request processing may be not pure parallel but fully asynchronous - I
placed some print statement in ajax controller and got full chaos - among
print from one request were one from another.
Here is my littlle test. 3 ajax calls start at once but finish
asynchronously (time.sleep()). By change time.sleep() it's possible to
get any order (I did "start first - finish last"). At the end of all
session contains only vars/changes of ajax which finish last.
import time
# it's a funny AJAX test
#just create new app and paste it in the default controller
# call .../default/many_ajx_form
def ajx_bug():
if request.args(0)=='0':
time.sleep(10) # well, now wait until fish fall asleep in the pond
session.req_0='Only req_0 ... nothing else!!! Where are req_1,
req_2 and other kids? It seems, that guys ignored my session lock?'
elif request.args(0)=='1':
time.sleep(5)
session.req_1='I killed req_2 and created req_1 which will be
killed by req_0 !'
else:
session.req_2='I created req_2'
response.flash="click 'Print Session' !"
if request.args(0)=='0':
return dict(ret='Better late than never')
else:
return dict(ret='completed req_%s'%request.args(0))
def print_session():
lst=[]
for k,v in session.items():
lst.append(DIV(
DIV(k,_class='col-xs-2', _style="font-weight:bold;
text-align:right; min-height:50px"),
DIV(':',_class='col-xs-1'),
DIV(v,_class='col-xs-9'),
_class='row'
)
)
return dict(r=DIV(*lst, _class="container", _style="max-width:800px"))
#------------ CALL THIS --------------------
def many_ajx_form():
session.clear() # - for clear test
session.dummy_data = 'dummy data' # make some action for sure init
session
ret={}
for i in xrange(3):
ret['ajx_frm_%s'%i]= \
DIV(
LOAD('default','ajx_bug.load',
args=[i],
ajax=True,
target='cont_%s'%i
),
_id='cont_%s'%i
)
btn_prnt=DIV(BUTTON("Print Session", _type="button",_class="btn
btn-default",
_onclick=myLOAD_any("default","print_session",
target="lst")),
DIV(_id="lst")
)
ret['z']=btn_prnt
return ret
#return string, no SCRIPT tag!
def myLOAD_any(contrl, fun, target, args=None, vars=None, ):
if vars:
vars_str='?'
for k,v in vars.items():
vars_str+=('%s=%s&'%(k,v))
vars_str= vars_str[:-1] # remove last '&'
else:
vars_str=''
if args:
args_str='/'.join([str(it) for it in args])
else:
args_str=''
data=dict(
rmt= "/%s/%s/%s.load/%s%s"%(request.application, contrl, fun,
args_str, vars_str ),
trg=target
)
load_str="""$.web2py.component("%(rmt)s", "%(trg)s", 0, 1,
$("#%(trg)s"));"""
return load_str%data
On Monday, August 10, 2015 at 4:05:57 AM UTC+3, Anthony wrote:
>
> All responses within the same session are supposed to have the same
> session_id. The fact that all three Ajax requests can access the session
> doesn't necessarily mean they are being processed in parallel (the requests
> might still be completing one after another).
>
> It would be helpful if you could attach a minimal app that demonstrates
> the problem.
>
> Anthony
>
> On Sunday, August 9, 2015 at 12:30:01 PM UTC-4, Val K wrote:
>>
>> I realize your advice - no effect!
>> I analyzed *session.connect* and found strange place at the biginig of
>> *connect *definition:
>> ...
>> self._unlock(response) # - *unconditional unlock *session file witch
>> have a name == response.session_id
>> ...
>>
>> Then I changed definition of ajx_bug(): form=SQLFORM.factory(Field('
>> any', *comment=response.session_id* ), table_name=form_name)
>> and here is I got:
>>
>> Many Ajx Form
>> ajx_frm_0:
>> Any
>> 127.0.0.1-dd2c6b6c-3141-4826-8887-ab1691b67350
>> ajx_frm_1:
>> Any
>> 127.0.0.1-dd2c6b6c-3141-4826-8887-ab1691b67350
>> ajx_frm_2:
>> Any
>> 127.0.0.1-dd2c6b6c-3141-4826-8887-ab1691b67350
>>
>>
>>
>>
>>
>>
>> All ajax responses have the same response_id! i.e. each ajax-process can
>> unlock session file locked by another!
>> In other words, requests parallel processing works properly across
>> sessions, but not within one session, because all responses within session
>> have the same response_id
>> It seems, that file is not locked across parallel process (on my Win7 at
>> least), may because it is children of one parent or something else?
>>
>>
>>
>> On Sunday, August 9, 2015 at 3:28:04 PM UTC+3, Anthony wrote:
>>>
>>> Does the problem occur only on the first page load of the session? If
>>> so, does the problem go away if you add the following line at the top of
>>> the many_ajx_form function:
>>>
>>> session.dummy_data = 'dummy data'
>>>
>>> If that's the case, then this is the same problem diagnosed in the
>>> original thread, with something like the above as the solution. The problem
>>> is not that Ajax reads fail to lock the session -- the problem is that at
>>> the very beginning of the session when the initial set of Ajax requests are
>>> made, there is no session file to lock at all. Adding some dummy data to
>>> the session in the parent page will force a session file to be created, so
>>> the session file will then exist and therefore be locked when the Ajax
>>> requests come in. Note, this won't work with sessions stored in the DB or
>>> in cookies, as there is no session locking at all with those methods (you
>>> will need some other means to avoid session race conditions in those cases).
>>>
>>> Anthony
>>>
>>>
>>>
>>>
>>>
>>>>
--
Resources:
- http://web2py.com
- http://web2py.com/book (Documentation)
- http://github.com/web2py/web2py (Source code)
- https://code.google.com/p/web2py/issues/list (Report Issues)
---
You received this message because you are subscribed to the Google Groups
"web2py-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to web2py+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
