This package enables server-side filtering, sorting and pagination of any IQueryable<T>
using OData syntax and without needing an EDM model.
This is achieved simply by adding [ODataQueryable]
to any such API.
Add a reference to Nuget package ODataQuery
.
// Import this namespace
using ODataQuery;
[Route("api/version")]
public class VersionController
{
// Add attribute ODataQueryable
[HttpGet, ODataQueryable]
public IQueryable<Version> Get()
{
// Source data can be anything (most likely an EF query)...
var versions = new[] {
new Version(1, 0),
new Version(1, 1),
new Version(2, 0),
new Version(3, 0),
};
// ... as long as it's IQueryable<T>
return versions.AsQueryable();
}
}
Now you can filter, sort and paginate versions:
GET /api/version?$filter=major gt 2&$orderby=major desc
would return this response:
{
"value": [
{ "major": 3, "minor": 0 },
{ "major": 2, "minor": 0 }
]
}
ODataQueryable
changes the shape of your response to a standard OData response:
{ "value": [...] }
by default;{ "@odata.count": 42, "value": [...] }
if you used the$count
option.
The goal of this project is to enable easy server-side processing of datagrids, lists and co. in web applications, while being secure by default
(e.g. you can't fetch the full DB through $expand
, because it's not supported).
Currently it has the following limitations:
- There is no routing, only query string processing.
- Not all system options are supported, only:
$filter
,$search
,$orderby
,$skip
,$top
,$select
and$count
. - Only a subset of the full OData 4.0.1 is supported, see below for detailed support.
- Error reporting is bad.
That last point warrants an explanation.
You will get a ParseException
if the query string is not correctly formatted and the message will indicate where it failed,
but the error message can often be unhelpful: e.g. mentioning a single expected character instead of a full token.
You will also get various exceptions if the query can be parsed but is semantically incorrect (properties that don't exist, type errors, invalid date literals).
In all cases this will likely result in a 500 Internal Error, although a 400 code (bad request) would be more appropriate.
With respect to error-friendliness, this API is not meant to be used to build public, open APIs.
Apply [ODataQueryable]
to automatically apply OData query string to an IQueryable<T>
result and return an OData response.
If the method throws, doesn't return a success code (2xx) or doesn't return an IQueryable<T>
result then this attribute does nothing.
This package adds an extension method to IQueryable<T>
to apply the query string manually in your code:
IQueryable<T> source;
// Does not apply $count
IQueryable<T> result = source.OData(HttpContext.Request.Query);
// Applies $count and return result in out parameter
int count;
IQueryable<T> result = source.OData(HttpContext.Request.Query, out count);
Two extension methods let you apply $filter
or $orderby
directly from a string:
IQueryable<T> source;
IQueryable<T> result = source.ODataFilter("amount gt 1000")
.ODataOrderBy("release desc,id");
This library follows OData conventions, with the following extensions:
-
DateTime
is supported (OData only supportsDateTimeOffset
). Literals without a timezone parse asDateTime
, literals with a timezone asDateTimeOffset
. -
in
operator can have an empty list on right hand side:x in ()
. This is forbidden by OData grammar but it is accepted by this library and evaluates tofalse
.
This library implements a subset of OData 4.0.1.
Only $filter
, $orderby
, $top
, $skip
and $count
are supported.
Supported: Numeric types, DateTime
and DateTimeOffset
, bool
, string
, enums.
Literals:
- strings delimited with single quotes, double single-quote escape:
'Marc''s house'
- numbers, exponents are not supported:
-? [0-9] (. [0-9] )?
- dates, day then optional time then optional timezone (Z or -0000), e.g.
2019-07-22
,2019-07-22T14:15:00
, and2019-07-22T14:15:00 0200
- enums, either as string or ints. Old syntax prefixed by type name is not supported:
'red'
. true
,false
null
Starting with a letter or _
, then followed by letters, digits and _
.
Identifier are looked for in a case-insensitive way in model. So id eq 3
will match a property Id
on the server.
Path to subproperties such as Address/City
are not supported.
Comparison operators:
- Supported:
eq
,ne
,lt
,le
,gt
,ge
,in
- Not supported:
has
Logical operators:
- Supported:
and
,or
,not
Arithmetic operators:
- Not supported:
add
,sub
,mul
,div
,divby
,mod
.
Grouping operators:
- Supported:
()
String functions:
- Supported:
contains
,endswith
,startswith
,length
,indexof
,substring
,tolower
,toupper
,trim
,concat
Date functions:
- Supported:
year
,month
,day
,hour
,minute
,second
,date
. - Not supported:
fractionalseconds
,time
,totaloffsetminutes
,now
,mindatetime
,maxdatetime
.
Math functions:
- Supported:
round
,floor
,ceiling
.
Type functions:
- Not supported:
cast
,isof
.
Geo function:
- Not supported:
geo.distance
,geo.length
,geo.intersects
.
Filter must evaluate to any boolean expression and is applied as Where
.
Collection predicates such as any
and all
are not supported, neither are paths to sub-properties such as Address/City
.
The semantics of what search does is up to the application: which fields? case-sensitive? starts, contains or exact match?
For this reason, the attribute ODataQueryable
simply binds a $search
query string to a search
parameter (if any), so that you can apply the search yourself.
Sorting is a comma-separated list of expressions, optionnally followed by keywords asc
and desc
.
It is applied as OrderBy
, ThenBy
, respectively OrderByDescending
, ThenByDescending
.
Note: OData grammar doesn't allow spaces before or after commas in $orderby
, neither does this library. So $orderby=name, id
will throw.
Projecting only supports a list of plain identifiers.
It is applied as a Select
that returns a lightweight IDictionary<string, object>
that is meant to be serialized (it is not fully functional and most methods throw NotImplementedException
).
If $count=true
is in the query string, the response will have an additional @odata.count
property set to the count of results just after applying filter, i.e. before applying $top
, $skip
, and $orderby
.
It is applied as a call to Count
(so 2 queries are executed).
Note: only the OData 4.0 $count
option is supported. Previously it was $inline-count=allpages
, which is not supported by this library.
Both supported as plain integers, applied as Take
and Skip
calls.
Note: before v1.1 this package looked for $take
instead of $top
which is not the OData standard.
For backward compatibility, $take
is still accepted as a synonym of $top
but it's usage is discouraged.