Re: Enforcing field type on publish


Benjamin Taieb
 

Hi Frank,

You right, I should specify further what we have in mind:

-Of course dictionary are not mandatory, and I don't think anyone wants the "dictionary check" to be default behaviour, it has to be optional to maintain compatibility with the many other use case. After all, mama is a middleware API, not a market data API and should be kept that way.

So property based behaviour here. In some respect, your point highlight also the fact that there is no way to prevent application developers from doing the "bad thing", as a property can always be changed. What we are after is to prevent applications developers from doing the "bad thing" by accident.

 

-UnitTest have been run and pass on our payload (well at least last time I checked), but last time I checked, they were no composite tests for scalar type for example.

The Qpid payload use a macro for all scalar, and in particular I believe:

pn_data_put##TYPE  (impl->mBody, value);

will result in "brutal" cast without any further check, i.e. truncation, interpreting double as I8 and all sort of nice things.

I don't think that is desirable from a design point of view, but would like to hear if people think there is a performance argument about it ?

Performance apart, I think also that defining behaviour in reference implementation is not desirable, comparing to having a specification. Just having a matrix somewhere (google spreadsheet or OpenMama wiki ?) will be very useful.

 

-That why the proposed changes is on the publisher, not on the subscriber, as in this case, there is strong asymmetry between publisher and subscriber, i.e. publisher are pretty low volume, but number of them, and subscribers will have higher traffic to process. I understand this is not the case for example with a low latency environment. That's why it is optional and not the default.

 

I agree about having flexibility in the cast belong to the payload, however I don't think UnitTest should verify the mandatory minimum, but rather enforces required interoperability, as this could makes payload incompatible. The whole benefit of using OpenMama collapse in this case in my opinion.

 

Cheers,

Ben.

 

 

From: Frank Quinn [mailto:fquinn@...]
Sent: 10 September 2014 09:31
To: Benjamin Taieb; Alpert, Reed; fquinn.ni@...; Glenn McClements
Cc: Openmama-dev@...
Subject: RE: [Openmama-dev] Enforcing field type on publish

 

Hi Ben,

 

There are a few things to consider here:

 

1.       The MAMA Dictionary is optional too. If you look at basic mamapublisherc / mamasubscriberc, you’ll see it doesn’t use the dictionary, but it does use mamaMsg.

2.       Unit tests actually do exist for the various mandatory castings required. If you run the payload unit tests today, for example, you’ll find that getting a date time from a string as well as a date time field should be supported, as well as other payload level casting expectations… In fact if you haven’t run these tests already, you really should, especially if you want to use MAMDA. If you look at the Qpid bridge, you’ll see a bunch of code put in to make this casting more forgiving.

3.       While iterating over a message, in the interest of performance, you probably don’t want to hit the data dictionary on every field being processed, particularly if you’re trying to do high performance field caching.

 

Basically, I’m of the opinion that if you want to make a payload flexible with respect to the field types it supports, that should be done at a payload layer and we have unit tests to verify the mandatory minimum.

 

If you want to enforce MAMA Dictionary compliance, this should be done in the MAMA application, where you can use field descriptors to verify content type at send and receive sides. E.g.

 

switch (mamaFieldDescriptor_getType (fieldDescriptor))

{

    case MAMA_FIELD_TYPE_I32:

    case MAMA_FIELD_TYPE_I64:

        allow the add / access;

    default:

        handle condition;

}

 

As well as everything else, whether or not the I64 to I32 casting (yes - the other way around) will be legal will depend on the data content being cast and sent across. Depending on the content, a U8 / U16 could even be fine. Without knowledge and consideration of this within the application, undesired truncation of data could be found, or opportunities to avoid failed data transfer due to strictness checking could be avoided.

 

Cheers,

Frank

 

From: openmama-dev-bounces@... [mailto:openmama-dev-bounces@...] On Behalf Of Benjamin Taieb
Sent: 10 September 2014 08:47
To: Alpert, Reed; fquinn.ni@...; Glenn McClements
Cc: Openmama-dev@...
Subject: Re: [Openmama-dev] Enforcing field type on publish

 

Just to put emphasis on what Reed explain:

It should be probably implemented in the C layer, making sure that it "catches" all languages "for free".

The goal is to enforce it in OpenMama layer, so that OpenMama application are forced to adheres to it.

 

I believe use of MamaFieldCache is optional, which would rules it out.

And referring to Glenn comments about mamaMsg, the goal is that the payload does not trust the calling application.

 

There is multiples challenges I believe in implementing that:

-How does the mama layer get dictionary ? This could be easily solves by using dictionnary requestor wherever we decides to implement it.

-How does the mapping between a particular msg and namespace is going to work ? At the moment, a dictionary is attached to namespace, namespace could be shared across multiples sources, but a message is source agnostic, only the publisher when it send the msg will send on a particular source/ symbol. On the other side, we would want to fail as soon as possible in the process, probably in the mamaMsg_addXX and mamaMsg_setXX calls, so that the calling application know for which fields it is doing something wrong.

-We need to clarify the "authorised cast": If you're doing addI32 on a I64 fields, should probably get authorised. At the moment, various Payload implement various level of flexibility, would be good that we defines UnitTests defining precisely what it should be, probably modelled from what is defined in wombatmsg, for retro-compatibility purpose.

 

I'll take any input from the community, particularly on point 2, which is the hardest in my opinion.

Cheers,

Ben.

 

From: openmama-dev-bounces@... [mailto:openmama-dev-bounces@...] On Behalf Of Alpert, Reed
Sent: 08 September 2014 20:10
To: fquinn.ni@...; Glenn McClements
Cc: Openmama-dev@...
Subject: Re: [Openmama-dev] Enforcing field type on publish

 

Yes, this is C++, but we also have Java and DotNet clients (some C# and some Excel).

We do publish best practices, but our app dev groups sometimes do not follow them, and we want to prevent the SolCache from getting out of sync with the dictionary.

We also provide them with sample apps that have this type of control, but sometimes … J

 

Thanks,

 

Reed.

 


Reed Alpert | Corporate & Investment Bank | Market Data Services | J.P. Morgan | 4 Metrotech Center, 23rd Floor, Brooklyn, NY 11245 | T: 718.242.5198  | M: 917.414.4613 | reed.alpert@...

 

Alternate Contact:  CIB PIM Trading Technology Solutions NA | CIB_PIM_Trading_Technology_Solutions_NA@...

 

From: Frank Quinn [mailto:fquinn.ni@...]
Sent: Thursday, September 04, 2014 4:56 PM
To: Glenn McClements
Cc: Alpert, Reed; Openmama-dev@...
Subject: Re: [Openmama-dev] Enforcing field type on publish

 

Guessing by your capitalization, I'm guessing this is C++? If so, have you considered mandating that all function calls use MamaFieldDescriptors pulled from a standardized dictionary rather than using string / fid pairs?

Cheers,
Frank

 

On Thu, Sep 4, 2014 at 9:33 PM, Glenn McClements <gmcclements@...> wrote:

There are a number of places where this *could* be done, but not all make sense to me:

 

- In the payload library - this doesn't feel right as I feel the payload should be 'dumb' and just trust the caller. Also it would need to be done for each supported payload. 

 

- In the payload bridge - same points as above

 

- In MamaMsg - this would work across all supported payloads, but again I think the payload should be trust the caller as to what fields should be what type and act a a simple wrapper 

 

- In a field cache (e.g. the C & C++ MamaFieldCache) - this makes the most sense to me as the logic for maintaining state about fields and if they have changed is already there, so extra rules and checks would belong here

 

With the current MamaFieldCache you could do this already if you primed the fields by creating them with the correct type on startup, this would depend on having a well known set of fields though because you don't want to be adding everything in the dictionary. You could also try this technique with the MamaMsg object, but it will be payload specific because some will throw an error if you try to update a field with a new type, and others may replace it.

 

Glenn 

 

 

 

From: Alpert, Reed [reed.alpert@...]
Sent: 03 September 2014 15:28
To: Glenn McClements; Openmama-dev@...


Subject: RE: Enforcing field type on publish

 

Hi,

 

Yes, the addString() is from MamaMsg.

 

The cache is our own market data infrastructure, not the MamaFieldCache.

In this case it is either Solace/SolCache or TREP RTIC.

 

We want business app dev groups to be able to look at the dictionary, see a field that is MamaPrice, and be guaranteed that MamaMsg.getPrice() will always work (although MamaPrice.getIsValidPrice() may return false).

A certain number of the type mismatches come from Excel, where a failed formula may populate a cell as #NA, and that gets published into a price/float field.

 

Thanks,

 

Reed.

 


Reed Alpert | Corporate & Investment Bank | Market Data Services | J.P. Morgan | 4 Metrotech Center, 23rd Floor, Brooklyn, NY 11245 | T: 718.242.5198  | M: 917.414.4613 | reed.alpert@...

 

Alternate Contact:  CIB PIM Trading Technology Solutions NA | CIB_PIM_Trading_Technology_Solutions_NA@...

 

From: Glenn McClements [mailto:gmcclements@...]
Sent: Wednesday, September 03, 2014 3:10 PM
To: Alpert, Reed; Openmama-dev@...
Subject: RE: Enforcing field type on publish

 

Hi Guy,

When you refer to 'addString' below you mention a cache, but this is not a method of the C/C++MamaFieldCache

are you refering to your own cache object or actually a MamaMsg object?

 

Regards,

Glenn 

 


From: openmama-dev-bounces@... [openmama-dev-bounces@...] on behalf of Alpert, Reed [reed.alpert@...]
Sent: 03 September 2014 13:59
To: Openmama-dev@...
Subject: [Openmama-dev] Enforcing field type on publish

Hi,

 

This is regarding a way to check that the field data type being published matches (closely) the dictionary data type.

 

The motivation for this is to prevent our internally published data caches from invalid data.

For example, if a field (name=MARK_PRICE fid=300) is a MamaPrice in the dictionary, and an app uses this:

                addString("MARK_PRICE", 300, "invalid price");

then we have a field that is now a string in the cache, and that causes problems with subscribing apps that use getPrice().

 

Currently we are looking to the bridges that we use (Tick42/TREP & Solace) to enforce these restrictions.

 

Adding this to the OpenMAMA layer is tricky, but allows a single policy to be enforced across multiple bridges. The main question from my point of view being how to set the policy. They can range from none (default), to most strict (published field's type must match dictionary), to common conversions supported (e.g., F32 -> F64), to a custom grid setting exactly what conversions are allowed.

 

I welcome any input from others who have run into this same issue, or found other ways to solve it.

 

Thanks,

 

Reed.

 


Reed Alpert | Corporate & Investment Bank | Market Data Services | J.P. Morgan | 4 Metrotech Center, 23rd Floor, Brooklyn, NY 11245 | T: 718.242.5198  | M: 917.414.4613 | reed.alpert@...

 

Alternate Contact:  CIB PIM Trading Technology Solutions NA | CIB_PIM_Trading_Technology_Solutions_NA@...

This communication is for informational purposes only. It is not intended as an offer or solicitation for the purchase or sale of any financial instrument or as an official confirmation of any transaction. All market prices, data and other information are not warranted as to completeness or accuracy and are subject to change without notice. Any comments or statements made herein do not necessarily reflect those of JPMorgan Chase & Co., its subsidiaries and affiliates. This transmission may contain information that is proprietary, privileged, confidential and/or exempt from disclosure under applicable law. If you are not the intended recipient, you are hereby notified that any disclosure, copying, distribution, or use of the information contained herein (including any reliance thereon) is STRICTLY PROHIBITED. Although this transmission and any attachments are believed to be free of any virus or other defect that might affect any computer system into which it is received and opened, it is the responsibility of the recipient to ensure that it is virus free and no responsibility is accepted by JPMorgan Chase & Co., its subsidiaries and affiliates, as applicable, for any loss or damage arising in any way from its use. If you received this transmission in error, please immediately contact the sender and destroy the material in its entirety, whether in electronic or hard copy format. Thank you. Please refer to http://www.jpmorgan.com/pages/disclosures for disclosures relating to European legal entities.


This message may contain confidential information and is intended for specific recipients unless explicitly noted otherwise. If you have reason to believe you are not an intended recipient of this message, please delete it and notify the sender. This message may not represent the opinion of Intercontinental Exchange, Inc. (ICE), Euronext or any of their subsidiaries or affiliates, and does not constitute a contract or guarantee. Unencrypted electronic mail is not secure and the recipient of this message is expected to provide safeguards from viruses and pursue alternate means of communication where privacy or a binding message is desired.


This communication is for informational purposes only. It is not intended as an offer or solicitation for the purchase or sale of any financial instrument or as an official confirmation of any transaction. All market prices, data and other information are not warranted as to completeness or accuracy and are subject to change without notice. Any comments or statements made herein do not necessarily reflect those of JPMorgan Chase & Co., its subsidiaries and affiliates. This transmission may contain information that is proprietary, privileged, confidential and/or exempt from disclosure under applicable law. If you are not the intended recipient, you are hereby notified that any disclosure, copying, distribution, or use of the information contained herein (including any reliance thereon) is STRICTLY PROHIBITED. Although this transmission and any attachments are believed to be free of any virus or other defect that might affect any computer system into which it is received and opened, it is the responsibility of the recipient to ensure that it is virus free and no responsibility is accepted by JPMorgan Chase & Co., its subsidiaries and affiliates, as applicable, for any loss or damage arising in any way from its use. If you received this transmission in error, please immediately contact the sender and destroy the material in its entirety, whether in electronic or hard copy format. Thank you. Please refer to http://www.jpmorgan.com/pages/disclosures for disclosures relating to European legal entities.


This message may contain confidential information and is intended for specific recipients unless explicitly noted otherwise. If you have reason to believe you are not an intended recipient of this message, please delete it and notify the sender. This message may not represent the opinion of Intercontinental Exchange, Inc. (ICE), Euronext or any of their subsidiaries or affiliates, and does not constitute a contract or guarantee. Unencrypted electronic mail is not secure and the recipient of this message is expected to provide safeguards from viruses and pursue alternate means of communication where privacy or a binding message is desired.



_______________________________________________
Openmama-dev mailing list
Openmama-dev@...
https://lists.openmama.org/mailman/listinfo/openmama-dev

 

This communication is for informational purposes only. It is not intended as an offer or solicitation for the purchase or sale of any financial instrument or as an official confirmation of any transaction. All market prices, data and other information are not warranted as to completeness or accuracy and are subject to change without notice. Any comments or statements made herein do not necessarily reflect those of JPMorgan Chase & Co., its subsidiaries and affiliates. This transmission may contain information that is proprietary, privileged, confidential and/or exempt from disclosure under applicable law. If you are not the intended recipient, you are hereby notified that any disclosure, copying, distribution, or use of the information contained herein (including any reliance thereon) is STRICTLY PROHIBITED. Although this transmission and any attachments are believed to be free of any virus or other defect that might affect any computer system into which it is received and opened, it is the responsibility of the recipient to ensure that it is virus free and no responsibility is accepted by JPMorgan Chase & Co., its subsidiaries and affiliates, as applicable, for any loss or damage arising in any way from its use. If you received this transmission in error, please immediately contact the sender and destroy the material in its entirety, whether in electronic or hard copy format. Thank you. Please refer to http://www.jpmorgan.com/pages/disclosures for disclosures relating to European legal entities.


This message may contain confidential information and is intended for specific recipients unless explicitly noted otherwise. If you have reason to believe you are not an intended recipient of this message, please delete it and notify the sender. This message may not represent the opinion of Intercontinental Exchange, Inc. (ICE), Euronext or any of their subsidiaries or affiliates, and does not constitute a contract or guarantee. Unencrypted electronic mail is not secure and the recipient of this message is expected to provide safeguards from viruses and pursue alternate means of communication where privacy or a binding message is desired.


Join Openmama-dev@lists.openmama.org to automatically receive all group messages.