the Token I used and exported is: eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJyb2xlIjoidG9kb191c2VyIn0.wj57GklEEYKxK5Av3qfhVY3sqWxRIvjGdTxcB0jcZGA
that's the entirety of it from jwt.io
am I supposed to use only a portion of the token?
@matrixbot SQL User Management : Did I set up according to the documentation
Hi there, I'm just discovering Postgrest. My team and I are evaluating it to replace one of our software. However, we're stuck on one functionality. Part of our model are big files. We obviously do not want to upload files into the DB but would it be possible to upload a file which would create an entry in the DB (id, updated date, creation date, path), and at the same time, save the file somewhere else (possibly another server, just acting as a storage filesystem).
Are these kind of scenario already been solved by someone else? Or is there some ideas about how this could be implemented with Postgrest?
Thanks in advance to anyone who would answer
Are these kind of scenario already been solved by someone else? Or is there some ideas about how this could be implemented with Postgrest?
Thanks in advance to anyone who would answer
(and by the way, Postgrest is really a nice product, even if in the end we do not use it, thanks to the contributors for this awesome idea!)
@woshilapin I haven't had to solve that exact problem, but I'd probably solve it the same way I've solved some related problems: intercept that endpoint with a Lambda, have the Lambda store the files in S3 (or retrieve them for GETs), and then send the metadata on to postgrest. (Swap out Lambda and S3 with whatever equivalents in your environment.)
My application is HTTPS-only, so for us 'interception' is pretty literal; I load-balance those endpoints to Lambda on the 443 listener and have a separate internal-only listener on 80 that the Lambdas send the metadata to; that way the autogenerated OpenAPI schema is still pretty-close-to-accurate. For other situations I imagine you'd just have to have separate routes and the OpenAPI schema would be less accurate.
The other approach would be to put the data into the database temporarily and
NOTIFY some process that pulls it out of the database and stores it elsewhere. The benefit of the lambda approach is that the file never even enters memory on the DB server, which I'd naively guess would help avoid performance impacts, but it really depends on how big the files are etc
This solution makes sense in AWS and I expect it makes sense in other cloud providers; if you're not running this on the cloud then I'm probably going to be less able to give advice :)
That might be helpful actually because we could just create the entry in the DB giving the URL where to download the file and notify an entry has been created. Then this process would read the entry and be able to download the file somewhere else, then update the DB with the final PATH...
ךושיל G² (Gitter): For JWT, you generally need two tokens: an access or identity token which has a short expiry and a refresh token which is usually an opaque blob that the client uses to get a new access token. The refresh token is stored in the database it can be checked and revoked. Have a look at subzero refresh token
I don't suppose there's a way to pass the secret base64 encoded?
Hello, I'm trying a simple joint
select abreviation from enseignants e
inner join people p
on e.id_people = p.id_people
where p.nom = 'CARAMBAR'
limit 1I write my route like this, but it doesn't affect the result, can you tell me where I'm wrong?
enseignants?select=*,people(*)&people.nom=eq.CARAMBAR&limit=1
@gerbyzation Have you tried using the secret-is-base64 config? In docker it should be
PGRST_SECRET_IS_BASE64=true.
select abreviation from enseignants e
left join people p
on e.id_people = p.id_people
where p.nom = 'CARAMBAR'
limit 1
I look at the logs to understand.
There's an equivalence of the SQL query and the pgrst request mentioned there.
@steve-chavez Okay, if I reverse the query and sort the result via my programming language, but I find it disturbing until I get used to it?
@MaitrePylos_gitlab Yeah, that's a limitation for now. When PostgREST/postgrest#1075 gets solved the requests will be more natural. Though left join will likely stay as default.
Hi all! I didn't know about this Gitter yesterday, so I posted an issue: PostgREST/postgrest#1422
Essentially, PostgreSQL doesn't allow SET LOCAL in STABLE functions. Is there a workaround? I'm trying to set the Content-Type of the response from within a function that I can call with a GET /rpc/get_blob?blob_slug=aslug. Thanks!
What you could do is to use to_number in a BEFORE INSERT TRIGGER.
Or perhaps just
replace could work:select replace('1234,56', ',', '.')::numeric;
@steve-chavez mmm, i'm looking for a way to accepts uploaded csv's in different formats, could be dutch, could be us, could be something else, depends on the user. How it's stored in the database isn't relevant, as long as can format it later in a way that suites the user. A trigger or replace would hardcode data upload to a specific format in the database...
@ako I see. In that case a VIEW or stored procedure would be better. You could upload the csv with an accompanying Content-Language header and then you could conditionally apply the transformation(specific to each lang). You can use GUCs to read the headers.
Hi all - massive apologies for being a total n00b here. Can I use LDAP for client-side authentication with PostgREST (without lots of work), and (2) I'm a bit of a moron, is there some nice person/people on here that might be able to answer my idiotic questions on that exact subject?
ךושיל Algy Taylor (Gitter): PostgREST uses JWT for authentication (and authorization to an extent) so you'll need to configure your LDAP service to return a JWT upon successful authentication. There seems to be a few LDAP JWT services.