Regression: ce.parse fails on expressions containing subscripts.

[rasmus-rasmus]

Description

I've been using ce.parse to parse latex expressions, where all variable names consist of one capital letter and one numerical subscript, e. g., 'A_1', 'B_2', 'Q_{22}', etc. These are preprocessed to the form 'A_\\text{1}' etc and then parsed.

This used to work, but as of v0.26.* expressions of this form are parsed to ['At', 'A', "'1'"] in stead giving errors when I'm trying to parse an expression such as for example 'A_1+B_1' since operand types are incompatible.

Steps to Reproduce

To just parse a variable name:

ce.parse('A_\\text{1}')

To parse an algebraic expression of several variables:

ce.parse('A_\\text{1} + B_\\text{1}')

Actual Result

First code snippet will return [ 'At', 'A', "'1'" ].

Second returns:

["At","A",["Add",["Error",["ErrorCode","'incompatible-type'","'number'","'string'"]],["Error",["ErrorCode","'incompatible-type'","'number'","'any'"]]]]

Expected Result

I would expect first to return ['A_1']
and second to return ['Add', 'A_1', B_1']

Environment

It worked in v0.25.1