Equivalent abstraction syntax yields different XMIR output

[nikololiahim]

The following two pieces of code should be syntactically equivalent: singleline.eo

(dir "/tmp").walk
  * ([f] (f.is-dir > @))

multiline.eo

walk.
  dir
    "/tmp"
  *
    [f]
      f.is-dir > @

However, the xmir produced by singleline.eo (this one is wrong, because .walk comes from (dir "string")): singleline.xml

...
<objects>
      <o base="dir" line="1" pos="1">
         <o base="string" data="string" line="1" pos="5">/tmp</o>
      </o>
      <o base=".walk" line="1" method="">
         <o base="array" data="array" line="2" pos="2">
            <o line="2" pos="5">
               <o line="2" pos="5">
                  <o line="2" name="f" pos="6"/>
                  <o base="f" line="2" pos="10"/>
                  <o base=".is-dir" line="2" method="" name="@"/>
               </o>
            </o>
         </o>
      </o>
   </objects>
...

is different from the xmir produced by multiline.eo (this one is correct)

...
<objects>
      <o base=".walk" line="1" pos="0">
         <o base="dir" line="2" pos="2">
            <o base="string" data="string" line="3" pos="4">/tmp</o>
         </o>
         <o base="array" data="array" line="4" pos="2">
            <o line="5" pos="4">
               <o line="5" name="f" pos="5"/>
               <o base="f" line="6" pos="6"/>
               <o base=".is-dir" line="6" method="" name="@"/>
            </o>
         </o>
      </o>
   </objects>
...

[yegor256]

@Graur please, verify this report (it's pretty old) and fix if the bug is still here

[Graur]

@l3r8yJ Please, take care of this

[l3r8yJ]

@Graur sure!

[l3r8yJ]

@Graur did I understand correctly that this class does not work correctly? If so, maybe we should transfer issue here?

[Graur]

@l3r8yJ For this ticket we don't care about any particular class. We need just to compare two xmir files produced from singleline EO representation and multiline

[l3r8yJ]

@Graur did i understand correctly that I need to reproduce the code in singleline and multiline, then compare xmir and say whether the bug report is relevant or not?

[Graur]

@l3r8yJ Absolutely right. But if they are different we need to fix it

[l3r8yJ]

@Graur got it!

[l3r8yJ]

@Graur have a look, please

Here is singleline:

<objects>
      <o abstract="" line="3" name="app" pos="0">
         <o line="3" name="args" pos="1" vararg=""/>
         <o alias="1" base="dir" line="4" pos="3">
            <o base="string" data="string" line="4" pos="7">/tmp</o>
         </o>
         <o base=".walk" line="4" method="" pos="14">
            <o base="array" data="array" line="5" pos="4">
               <o alias="2" line="5" pos="7">
                  <o abstract="" line="5" pos="7">
                     <o line="5" name="f" pos="8"/>
                     <o alias="3" base="f" line="5" pos="12"/>
                     <o base=".is-dir" line="5" method="" name="@" pos="13"/>
                  </o>
               </o>
            </o>
         </o>
      </o>
</objects>

Here is multiline:

<objects>
      <o abstract="" line="3" name="app" pos="0">
         <o line="3" name="args" pos="1" vararg=""/>
         <o base=".walk" line="4" pos="2">
            <o base="dir" line="5" pos="4">
               <o base="string" data="string" line="6" pos="6">/tmp</o>
            </o>
            <o base="array" data="array" line="7" pos="4">
               <o abstract="" line="8" pos="6">
                  <o line="8" name="f" pos="7"/>
                  <o base="f" line="9" pos="8"/>
                  <o base=".is-dir" line="9" method="" name="@" pos="9"/>
               </o>
            </o>
         </o>
      </o>
</objects>

in multiline we're good, but in singleline we get an error: The object has attribute without a name, line=5, pos=7 (noname-attributes:22)

btw, code example was taken from a pdf about eo:

[l3r8yJ]

@Graur have a look, please Looks like

(dir "/tmp").walk
  [f] f.is-dir > @

works fine, we don't need the * symbol before [f].

also xmir is the same as multiline:

<objects>
      <o abstract="" line="3" name="app" pos="0">
         <o line="3" name="args" pos="1" vararg=""/>
         <o alias="1" base="dir" line="4" pos="3">
            <o base="string" data="string" line="4" pos="7">/tmp</o>
         </o>
         <o base=".walk" line="4" method="" pos="14">
            <o abstract="" line="5" pos="4">
               <o line="5" name="f" pos="5"/>
               <o base="f" line="5" pos="8"/>
               <o base=".is-dir" line="5" method="" name="@" pos="9"/>
            </o>
         </o>
      </o>
</objects>

Where do I need to create an eo pdf fixing issue?

[Graur]

@l3r8yJ You can find it here: https://github.com/objectionary/eo/blob/master/paper/sections/syntax.tex#L480-L506. And commit your changes just like you do it with code

[l3r8yJ]

@nikololiahim @yegor256 have a look, please

I compared

walk.
  dir
    "/tmp"
  [f]
    f.is-dir > @  

and

(dir "/tmp").walk
  [f] > (f.is-dir > @)

here is a comparison between the correct variants multiline:

   <objects>
      <o abstract="" line="3" name="app" pos="0">
         <o line="3" name="args" pos="1" vararg=""/>
         <o base=".walk" line="4" pos="2">
            <o base="dir" line="5" pos="4">
               <o base="string" data="string" line="6" pos="6">/tmp</o>
            </o>
            <o abstract="" line="7" pos="4">
               <o line="7" name="f" pos="5"/>
               <o base="f" line="8" pos="6"/>
               <o base=".is-dir" line="8" method="" name="@" pos="7"/>
            </o>
         </o>
      </o>
   </objects>

singleline:

   <objects>
      <o abstract="" line="3" name="app" pos="0">
         <o line="3" name="args" pos="1" vararg=""/>
         <o alias="1" base="dir" line="4" pos="3">
            <o base="string" data="string" line="4" pos="7">/tmp</o>
         </o>
         <o base=".walk" line="4" method="" pos="14">
            <o abstract="" line="5" pos="4">
               <o line="5" name="f" pos="5"/>
               <o alias="2" base="f" line="5" pos="9"/>
               <o base=".is-dir" line="5" method="" name="@" pos="10"/>
            </o>
         </o>
      </o>
   </objects>

In comparsion, are we looking to base tag? If so, it looks like the same xmirs now, doesn't it?

[l3r8yJ]

@yegor256 please take a look, it looks like the author of the issue is not active, what should we do?

[yegor256]

@mximp please, help us here

[mximp]

@l3r8yJ I still see some issues. Have you tried to run the single-line code:

(dir "/tmp").walk
  [f] > (f.is-dir > @)

I suppose it will fail as according to XMIR it tries to call .walk on the abstract object [f] > (f.is-dir > @) whereas it is expected to be called on (dir "/tmp")

[yegor256]

@mximp where we are with this problem?

[l3r8yJ]

@mximp I went back to this issue and will solve it

[maxonfjvipon]

@l3r8yJ where are you with this problem? If you don't mind I can take it

[l3r8yJ]

@maxonfjvipon sure, I don't have time to fit it into my schedule ; (

[maxonfjvipon]

@yegor256 @Graur I don't see any problem here. I simplified the example:

if. > first
  TRUE
  1
  2
TRUE.if > second
  1
  2

This is xmir after PARSING step and yes, these objects look differently:

<o base=".if" line="2" name="first" pos="2">
<o base="bool" data="bytes" line="3" pos="4">01</o>
    <o base="int" data="bytes" line="4" pos="4">00 00 00 00 00 00 00 01</o>
    <o base="int" data="bytes" line="5" pos="4">00 00 00 00 00 00 00 02</o>
</o>
<o base="bool" data="bytes" line="6" pos="2">01</o>
<o base=".if" line="6" method="" name="second" pos="6">
    <o base="int" data="bytes" line="7" pos="4">00 00 00 00 00 00 00 01</o>
    <o base="int" data="bytes" line="8" pos="4">00 00 00 00 00 00 00 02</o>
</o>

But they become the same after OPTIMIZATION step:

<o base=".if" line="2" loc="Φ.main.first" name="first" pos="2">
    <o base="org.eolang.bool"
       data="bytes"
       line="3"
       loc="Φ.main.first.ρ"
       pos="4">01</o>
    <o base="org.eolang.int"
       data="bytes"
       line="4"
       loc="Φ.main.first.α0"
       pos="4">00 00 00 00 00 00 00 01</o>
    <o base="org.eolang.int"
       data="bytes"
       line="5"
       loc="Φ.main.first.α1"
       pos="4">00 00 00 00 00 00 00 02</o>
</o>
<o base=".if" line="6" loc="Φ.main.second" name="second" pos="6">
    <o base="org.eolang.bool"
       data="bytes"
       line="6"
       loc="Φ.main.second.ρ"
       pos="2">01</o>
    <o base="org.eolang.int"
       data="bytes"
       line="7"
       loc="Φ.main.second.α0"
       pos="4">00 00 00 00 00 00 00 01</o>
    <o base="org.eolang.int"
       data="bytes"
       line="8"
       loc="Φ.main.second.α1"
       pos="4">00 00 00 00 00 00 00 02</o>
</o>

I believe there's no dogma that the same objects written in different notations should look the same after parsing step. That's why I think we can just close the ticket

[yegor256]

@Graur I agree