Re: Semi-unable to add new records to table--primary key needed?
От | Boylan, Ross |
---|---|
Тема | Re: Semi-unable to add new records to table--primary key needed? |
Дата | |
Msg-id | BYAPR05MB573627502612B1A2EE3CC33D872F0@BYAPR05MB5736.namprd05.prod.outlook.com обсуждение исходный текст |
Ответ на | Re: Semi-unable to add new records to table--primary key needed? (Adrian Klaver <adrian.klaver@aklaver.com>) |
Ответы |
Re: Semi-unable to add new records to table--primary key needed?
|
Список | pgsql-general |
> From: Adrian Klaver <adrian.klaver@aklaver.com> > Sent: Sunday, December 22, 2019 10:35 AM ..... > Alright this is the part where I got confused. I think what is going on is: >1) The immediate change is going to be to Access 2016 on Windows 10 64 keeping the data in Access files(.accdb) > 2) The long term plan is to move the data from the Access files to a stand alone SQL server, presumably Postgres. You are looking for assistance in converting Access tables and relationships to Postgres. > 3) At some point the GUI will point to the data in the server instead of in the Access files. > Is any of the above correct? Basically, yes. 1) The current file is in the older Access .mdb format, and will stay that way for the immediate conversion to Win 10/Access2016. 2) I'd like to go to Postgres medium term. I have converted the Access data to Postgres via an Access/VBA program that writesout SQL files. It clearly needs some work. I think I may have everything I need to tweak it. 3) Basically yes, though in detail the GUI and the code behind it directly use "linked tables" that are local to the frontend. These linked tables are the Access equivalent of symlinks to a particular table on a particular database. Theywill use the PG ODBC driver to do the talking. The app uses those linked tables now, but the link simply points to thebackend file on the disk. One other issue came up in testing: searches against the Access backend are case insensitive, while searches against thePG backend are case-sensitive. It would be nice to make them case-insensitive, and I've been reading about PG collationoptions to do so. Case-sensitivity is completely distinct from all the other issues I've discussed. The sense was also that the problem/differencewas not a show-stopper. Current behavior is that if you search for id abc123 it will retrieve an id storedas ABC123. With the PG backend, searching on abc123 retrieves nothing. > More below. >>> 3) Have you looked at the Relations tab in Access to see what if any relationships are there? .... > >> Yes, but the export program doesn't :) The relations tab documents many, but not all, of the relations in the database. The relations are also a little tricky because sometimes the lack of a relation should not be considered disqualifyingfor a specimen. Simple example: freezer type is an id to be looked up in a small table of freezer type idsand their names. If the freezer type is missing or nonsense, we may still want the sample. That can be expressed asa left join; the "Access SQL is not SQL" problems centered on left joins. >Postgres is going to be stricter about this. Access has the concept of suggested relationships that are not enforced: >https://urldefense.proofpoint.com/v2/url?u=https-3A__support.office.com_en-2Dus_article_Guide-2Dto-2Dtable-2Drelationships-2D30446197-2D4fbe-2D457b-2Db992-2D2f6fb812b58f&d=DwIDaQ&c=iORugZls2LlYyCAZRB3XLg&r=nh70E5-mX2XsDe5lrDDMt_ZRqqGMcdTmTrRLvNmttYA&m=qa1d513KOtyWtV0u2K81rduJ0TiRaucF3gbnjVTESkc&s=F10pVnItttte6dDN-4R4t13xDrJr8zAoIp63mrYNMh0&e= Thanks for the pointer. I'm not sure if we're using any of those suggested relationships, but we're definitely using relationshipsthat one infers only by looking at the SQL code. In other words, queries link tables by fields that are notmentioned on the relationships tab. ...... > Postgres Foreign Key relationships either exist or they don't. Now you can employee 'cheats' with you own triggers, but that is another subject. >As to your freezer example: >create table freezer_info(freezer_id int PRIMARY KEY); >create table sample_tbl(sample_id serial PRIMARY KEY, freezer_fk int >REFERENCES freezer_info ON UPDATE CASCADE); The direct translation of our current app would be to omit the foreign key relation in the table definition, but to havequeries that include a left join from sample_tbl to freezer_info. That raises one other question: what is the relation between the relations expressed on the server via FOREIGN KEY declarationsand relations known to Access? At first blush, they are completely different. In Access a relation is an objectthat lives on the front-end, while the server relations obviously live on the server. I don't know if Access, perhapsvia the ODBC spec, does anything to guarantee their consistency. Clearly since Access relations can have "suggestedrelations" which have no backend equivalent, the 2 sets of relations cannot in general be mirror images. Identifying and enforcing all foreign key relations, is something I'm planning to defer til after the migration to PG. Thereis a noticeable amount of funky stuff in the data, which could use some housekeeping regardless of platform. The cleanupis not likely to be simple. One of the long-run benefits of using a database server is that it should limit the possibilities for funkiness by enforcingreferential integrity and properly cleaning up after incomplete operations, instead of leaving pieces of them inthe database. But we can't enforce referential integrity until our data exhibit it! And the transactional integrity probablydepends more on proper coding on the front-end app than the choice of backend; file-based Access backends do supporttransactions. Ross
В списке pgsql-general по дате отправления:
Следующее
От: Adrian KlaverДата:
Сообщение: Re: Semi-unable to add new records to table--primary key needed?