create new field based on multiple resident tables - join

Question

Given multiple in-resident tables, I'd like to create a new field based on fields in different tables.
table1:
LOAD * INLINE [
id1,val1
a1,car1
a2,car1
];
table2:
LOAD * INLINE [
id2,id1,val2
b1,a1,type1
b2,a2,type2
];
table3:
LOAD * INLINE [
id3,id2,val3
c1,b1,mfr1
c2,b2,mfr2
];
For the sake of argument, assume table1 has ~1M rows, table2 ~1K rows, and table3 ~10 rows. I'd like to create a new field that is either added to table1 or perhaps in a new table linked by id1, resulting in:
id1 val1 newval
a1 car1 car1type1mfr1
a2 car2 car2type2mfr2
Efforts:
newtable:
load val1 & val2 & val3 as newval;
No errors but no newtable or newval.
newtable:
left join (table2)
load val1&val2 as newval resident table1;
Errs with Field not found - <val2>. (Obviously I want to extend this to include table3, but if I can't do it with 2 tables then 3 just won't work.
The real data includes seven tables for this new field (lots of foreign keys). The data is being loaded from QVDs (the data is shared across multiple QVWs), closely mimicking a SQL database; none of the tables are row-wise redundant, so combining db tables into a single QVD table may be inefficient. (Plus refreshing the data is incredibly easier one table at a time.) A colleague suggested I load-join each of the QVDs into one huge table, but that doesn't seem right (nor have I successfully chain-joined even a few tables).
Using QV 12.0 desktop on win10-x64 for deployment on QVS.

Answer 1

#TheBudac's was part of the way there, but it only merged two of the three. Most of the problems were stemming from incorrect multi-table joins. My confusion was in the "join" syntax in Qlik; the docs make sense to me now that I see what's happening, but it wasn't as obvious to me initially.
Here's what eventually worked best for me:
temptable:
load id1 as id1a, val1 as val1a
resident table1;
left join (temptable)
load id2 as id2a, id1 as id1a, val2 as val2a
resident table2;
left join (temptable)
load id2 as id2a, val3 as val3a
resident table3;
newtable:
load id1a as id1,
val1a & val2a & val3a as newval
resident temptable;
drop table temptable;
This produced these tables:
and this tree:
Quick walk-through:
Because I'm using left join, I start with the largest table; other joins would dictate different starting condition requirements. In my case, table1 was representing the largest, so I start with that:
temptable:
load id1 as id1a, val1 as val1a
resident table1;
Each join should be against the temporary table we're working on. Renaming variables is important so that Qlik doesn't create unnecessary synthetic keys.
left join (temptable)
load id2 as id2a, id1 as id1a, val2 as val2a
resident table2;
The use of resident is important in that it does not re-query (SQL) or re-load (QVD or other file).
Repeat with the third and further tables, always joining against temptable with the new table.
Now we use that temporary table to create our new table. You can choose to augment table1 with this data instead (certainly feasible), but for me since I'm generating several new calculated fields (not shown here), it made sense to keep them logically separated.
newtable:
load id1a as id1,
val1a & val2a & val3a as newval
resident temptable;
drop table temptable;
Note that I rename the relevant key back to its original value so that this table correctly links to table. Dropping the temporary table helps clean things up, but it does no harm to keep it around (and doing so helps in debugging/learning).

Answer 2

Your join is the wrong way round and QlikView can only work results after they have been joined,not in process, so you will have to do another resident load to get the values concatenated into Newval. The drop table commands are important or you will end up with massive unintentional syn tables
newtable:
left join (table1)
load * resident table2; drop table 2;
Resulttable:
load id1,
val1&val2 as NewVal
resident newtable; drop newtable;