The header comment on CKFetchDatabaseChangesOperation fetchDatabaseChangesCompletionBlock states:
"If the server returns a CKErrorChangeTokenExpired error, the previousServerChangeToken value was too old and the client should toss its local cache and
re-fetch the changes in this record zone starting with a nil previousServerChangeToken."
I would like to test this scenario thus I would like to generate an expired CKServerChangeToken so I can set it as the previousServerChangeToken on a CKFetchDatabaseChangesOperation.
I added an init method from the private header:
#interface CKServerChangeToken (Private)
- (id)initWithData:(NSData *)data;
#end
And used it as follows:
CKServerChangeToken knownExpiredToken = [[CKServerChangeToken alloc] initWithData:[[NSData alloc] initWithBase64EncodedString:#"AQAAAVl57tUGHv6sgNT9EeaTcQCM+sDHHA==" options:0]];
That string is a valid change token returned from a request and I have tried unsuccessfully modifying it, e.g. reducing numbers that I see incrementing to lower ones. I have however managed to get another strange invalid argument errors like continuation marker missing. I would be grateful if a CloudKit engineer has any suggestions, thanks.
This is an old post, but I've unintentionally generated .changeTokenExpired errors by doing this:
Fetch changes and save your change token.
In the dashboard, delete the zone you're syncing with.
Create a new zone with the same name.
In your code, fetch changes again, using the change token from step 1.
Since the token refers to a different zone, it doesn't make any sense to CloudKit, which returns a .changeTokenExpired error.
It would be really nice if Apple provided a way to do this, because it's an absolutely crucial situation to test.
I couldn't find any recommended way to do this, so I used tokens generated for other zones. If you've got multiple zones like I did, you can effectively cross wire token retrieval. It's not an elegant solution, but it does work... :)
Related
I have a datetime value which comes from the API in this format: 2015-07-07T17:30:00+00:00. I simply want to split it up between the date and time values at this point. I am not using an Active Record model and I prefer not to use an sql database if I can.
The way I have set up the app means that the value is "stored" like this in my view: #search.dining_date_and_time
I have tried two approaches to solving this problem:
Manually based on this previous stackoverflow question from 2012: Using multiple input fields for one attribute - but the error I get is the attribute is "nil" even though I put a "try"
Using this gem, https://github.com/ccallebs/split_date_time which is a bit more recent and seems to be a more elegant solution, but after closely following the doc, I get this error, saying my Search model is not initalized and there is no method: undefined method dining_date' for #<Search not initialized>
This is when instead I put #search.dining_date in the view, which seems to be the equivalent of the doc's example (its not that clear). The doc also says the method will be automatically generated.
Do I need to alter my model so I receive the data from the API in another way? ie. not get the variable back as #search.dining_date_and_time from the Search model for any of this to work?
Do I need an Active Record model so that before_filter or before_save logic works - so i can (re)concatenate after splitting so the data is sent back to the API in a format it understands. Can I avoid this - it seems a bit of overkill to restructure the whole app and put in a full database just so I can split and join date/time as needed.
Happy to provide further details, code snippets if required.
As I am not using a conventional Rails DB like MySql Lite or Postgresql, I found that the best solution to the problem was by using this jQuery date Format plugin: https://github.com/phstc/jquery-dateFormat to split the date and time values for display when I get the data back from the API.
The Github docs were not too expansive, but once I put the simply put the library file in my Rails javascript assets folder, I just had to write a few lines of jQuery to get the result and format I wanted:
$(function() {
var rawDateTime = $('#searchDiningDateTime').html();
// console.log(rawDateTime);
var cleanDate = $.format.date(rawDateTime, "ddd, dd/MM/yyyy");
// console.log(cleanDate);
$('#searchDiningDateTime').html(cleanDate);
var cleanTime = $.format.date(rawDateTime, "HH:mm");
// console.log(cleanTime);
$('#searchTime').html(cleanTime);
});
Next challenge: rejoin the values on submit, so the API can read the data by sending/receiving a valid request/response. (The values can't be split like this when sent to the remote service).
From the research I've done, it appears I need to send a special OID with my request (1.2.840.113556.1.4.417) in order to access the Deleted Objects container.
I couldn't find a way to send a specific control with a request using the "net-ldap" gem. Does anyone know if this is possible?
There is another gem, ruby-ldap, which appears to be more flexible and it seems I can send controls with my request (e.g. using the search_ext2() method).
However, no matter what I try, I am not getting back any objects, even though I know they haven't been garbage collected yet.
I'm including the filter "isDeleted=TRUE" with my requests as well.
OK, I finally figured it out. One will need to use the ruby-ldap gem. The reason my controls were not being sent was because the LDAP Protocol Version (LDAP::LDAP_OPT_PROTOCOL_VERSION) had defaulted to v2 and apparently it must be v3.
The following is a snippet that works:
require 'ldap'
conn = LDAP::Conn.new('yourserver.example.com', 389)
conn.set_option(LDAP::LDAP_OPT_PROTOCOL_VERSION, 3)
conn.bind("CN=Administrator,CN=Users,DC=example,DC=com", "sekritpass")
# controlType: 1.2.840.113556.1.4.417 (LDAP_SERVER_SHOW_DELETED_OID)
control = LDAP::Control.new('1.2.840.113556.1.4.417')
conn.search_ext2('CN=Deleted Objects,DC=example,DC=com', LDAP::LDAP_SCOPE_SUBTREE, "(isDeleted=*)", nil, false, [control], nil)
The filter (isDeleted=*) isn't necessarily required, you could also simply use (objectClass=*). You can also use the scope LDAP::LDAP_SCOPE_ONELEVEL if desired.
Have you tried isDeleted=* instead?
https://technet.microsoft.com/en-us/library/cc978013.aspx
It would appear that the savechanges method on breeze waits indefinitely when calling to or waiting for the server. Is there a way of getting it to time out? I am calling save change with allowConcurrentSaves: false. This now causes users who somehow do not get a response from the server to simply hang in limbo indefinitely say for example with a dropped internet connection.
I do not want to re-call the method with allowConcurrentSaves to false fearing that I might duplicate the data.
Any ideas?
Thanks
Update 16 May 2014
You can set HTTP-level timeout and cancellation with the AJAX Adapter's requestInterceptor as of v.1.4.12. See the documentation, "Controlling AJAX calls".
I'd still be reluctant to use this feature on save as you have no chance of knowing what whether the server persisted the data or not. Of course if your client hangs or crashes you don't know anyway. It's up to you.
Original Answer
Actually, there is a ready-made solution from Q.js. It's called timeout and it's mentioned in the API reference with a simplified example of its implementation and use in the readme.md.
I know you asked about Save but your question is pertinent for promises in general. Here is a query example adapted from the queryTests.js in our DocCode Sample
var timeoutMs = 10000; // 10 second timeout
var em = newEm(); // creates a new EntityManager
var query = new EntityQuery().from("Customers").using(em);
Q.timeout(query.execute, timeoutMs)
.then(queryFinishedBeforeTimeout)
.fail(queryFailedOrTimedout);
function queryFailedOrTimedout(error) {
var expect = /timed out/i;
var emsg = error.message;
if (expect.test(emsg)) {
log("Query timed out w/ message '{0}' " + expectTimeoutMsg)
.format(emsg));
// do something
} else {
handleFail(error);
}
}
Note: I just added this test so you'd have to get if from github or wait for a Breeze release after 1.2.5.
Oops ... maybe not
I gave what I think is a great answer for query. It may not be the right answer for save.
The problem with save is that you do not know on the client if the save succeeded until the server responds. Things could go wrong anywhere along the way. The server might not have heard the request to save. The server may have failed during save. The server may have saved the data but the response never made it back to the client.
Changing the value of allowConcurrentSaves won't get you out of this bind. Neither will having a save timeout.
In fact, adding a timeout to the save is probably deceiving. It is even possible for the save response to arrive after your custom timeout ... in which case Breeze will have tried to update your EntityManager ... and you won't know if Breeze succeeded or failed!
What if we added a Breeze save timeout. What should it do? What if breeze said the save had timedout ... and Breeze ignored a belated response from the server? Then imagine that the save succeeded on the server - it just took "too long" for it to respond to the client. Now you've got a client whose state is unexpectedly out of sync with the server. This is not good.
So I think you want a different solution to this very real problem. It's a user experience problem really. You can indicate to the user that you think the save is still in progress and then set your own timer. If the save isn't done when your timer expires, you can query the server to see if the data have been saved or if there is a connection ... or something along these lines. I can't think of a better way right now honestly.
Note that I'm assuming you need to know that the server succeeded. If you avoid store-generated IDs and always assume saves succeed unless the server tells you otherwise ... well that's a completely different paradigm and programming model that we could talk about someday (see meteorjs).
The net of all of this: I'm pretty darned sure that a save timeout is NOT what you want.
Still useful on a query though :)
Great question, and I wish I had a good answer. But it is definitely worth looking into. Could you please add this as a feature request to the Breeze User Voice. We take these requests very seriously in determining our priorities for Breeze development.
First time user long time reader. I have thoroughly looked for an explanation for the problem I'm having via the mighty search engine Google, but alas I have failed to produce any significant insight.
I need to be able to ensure that a model form is not reloaded with invalid data. Since the model stored in memory on the server is edited directly with the parameters of the web form first, and THEN checked for validity, without additional code invalid model data will ALWAYS be sent back to the form. This is less than desirable to me. My question is this: how do I ensure this doesn't happen?
What I'm thinking is I need some mechanism for saving the state of the object before it's modified with the parameters sent from the web form, and then after a failed validation restore the object to it's previous, correct and unmodified state of being.
Help!
Thanks,
Les
The object isn't actually modified in the db if validation fails, even though the object is in an invalid state in the form ... the thinking behind this is that the user wants to see the errors they made so they can correct them.
If you don't want that to be the case, then just read back the object with a WhateverObject.find(x) and assign it to the variable that the form is referencing and it will 'restore' the object to its previous unmodified state.
To add to what concept47 said you can also get the value for a particular field using
object.field_was
Have a look at ActiveRecord::Dirty for details (http://ar.rubyonrails.org/classes/ActiveRecord/Dirty.html)
Using that you could retrieve the original values for just those fields that had validation errors.
In our grails application we're logging a lot, but need a mechanism to associate all of those messages with the request/response being processed. It has proven easy enough to generate a request UUID, but now I'd like that id appended to each log message generated within a request context without passing that id within each log message. Has anybody implemented such a system so that you can associate all of your log statements together?
A rather obscure feature of log4j, called MDC seems to be exactly what you need.
Something like http://gustlik.wordpress.com/2008/07/05/user-context-tracking-in-log4j/
It will work fine in Grails as well if you use a custom AppFilter to set the request-unique value.
you could try utilizing the RequestContextHolder
http://static.springsource.org/spring/docs/3.0.x/javadoc-api/org/springframework/web/context/request/RequestContextHolder.html
ServletRequestAttributes attr = (ServletRequestAttributes) RequestContextHolder.currentRequestAttributes();
log.debug attr.getRequest().getSession()
Once you get the session object, you can get whatever identifier you have stashed away?